In a few weeks, someone new will walk through the most famous black door in the country.

The cameras will flash. The speech will be made. The red boxes will start to arrive.

As things stand, that person is Andy Burnham.

Subscribe now

He won the Makerfield by election a week ago, on 18 June, with a majority of more than nine thousand.

Four days later, Keir Starmer quit.

Nominations open on 9 July. His main rival has stepped aside to back him. Barring a shock, he is our next Prime Minister.

He will inherit the usual things. A shaky economy. A split party. A country that feels like it has stopped working.

But one file in that red box is bigger than all of it. And almost nobody is talking about it.

The NHS. Defence. Tax. Policing.

We do not own the infrastructure our own state runs on. And we have no real plan for what happens if that ever becomes a problem.

Let me show you how deep this goes.

We are spending billions on tech we do not own

Start with one number.

The UK public sector spends around £16.6 billion a year with technology suppliers. Most of it goes to American firms.

Two companies, Amazon and Microsoft, each control between 30 and 40 percent of the British cloud market.

So what does this mean?

Because “the cloud” sounds vague and harmless. Like weather.

It is not.

The cloud is just someone else’s computer.

Row after row of machines, in a warehouse, owned by a company, running the software that keeps the country going.

When the NHS holds your medical records, when HMRC handles your tax, when the Home Office tracks a visa, that data lives on machines owned by American firms.

And the deals keep growing. On a single day in December 2023, Amazon signed three contracts worth £894 million with the Home Office, HMRC and the Department for Work and Pensions.

Amazon made more from those three deals in one day than it had in its entire previous decade of selling cloud to the British government.

And it is speeding up. This April, HMRC handed Amazon another £473 million to move our tax system onto its servers for the next ten years.

When the work went out to the market, every other bidder pulled out. Amazon was the only one left.

Microsoft takes roughly £1.9 billion a year in government licences. The NHS is moving 1.5 million staff onto Microsoft 365.

We once had our own British provider, UKCloud. It went bust in October 2022. Nobody replaced it.

And here is the bit that should embarrass us. In March this year, Parliament’s own research service confirmed it in black and white. The UK has no overarching policy on digital sovereignty. None.

The kill switch is not a theory. It already happened.

This is not hypothetical. We have seen it.

In February 2025, Donald Trump sanctioned the chief prosecutor of the International Criminal Court.

Within months, his Microsoft email stopped working.

A court investigating war crimes was locked out of its own systems. Because an American company was doing what American law told it to do.

Read that again, slowly.

The dependency itself was the weapon.

Then came Anthropic, the American AI company. It refused to let the Pentagon use its technology for fully autonomous weapons.

So the US government branded it a “supply chain risk”. That is a label normally saved for hostile foreign powers.

In March, a judge threw the label out and called it Orwellian.

So here is the uncomfortable question. If Washington will do that to one of its own companies, over a disagreement, what protects Britain?

Nothing.

There is a piece of American law called the CLOUD Act, passed in 2018. In plain English, it lets American authorities force American companies to hand over data wherever in the world it sits.

An order served in Virginia can reach a server in London.

A Microsoft boss admitted exactly this under oath in France last year. They could not, they said, promise otherwise.

This is the inheritance. Not a partnership. A dependency. And dependency is just another word for leverage.

The Palantir problem lands on PM’s desk first.

Palantir is a surveillance company co founded by Trump’s billionaire ally Peter Thiel. It now holds at least £670 million in UK contracts. Add the defence deal from December and the total climbs past £900 million.

That defence contract, worth £240.6 million, was handed over with no competition at all.

It is meant to be the exception, not the rule.

Worse, it was signed soon after Palantir had hired a string of senior defence officials, including people who had worked on Britain’s own military AI strategy. One reportedly left the ministry and joined the company within days.

This is not a conspiracy theory. It was documented by The Nerve, run by Carole Cadwalladr, and by openDemocracy. It is on the public record.

Then there is the NHS. Palantir won a contract worth up to £330 million to run a platform that pulls together patient data across the health service.

The British Medical Association, the doctors’ union, has told its members to steer clear.

A break clause, a get out option in the contract, opens in 2027.

And here is the detail that makes this Burnham’s fight in particular.

Greater Manchester said no.

Of the 42 regional NHS bodies in England, his own patch is the one that refused Palantir and built its own system instead. One its own data chief calls years ahead of what Palantir offered.

His own backyard already proved there is a better way.

So the question writes itself. Will the man who promised to “end 40 years of neoliberalism” cancel the contracts handed to Peter Thiel?

Or does the bold talk stop at the door of the people who quietly own our data?

Europe chose differently. We can too.

The good news is that none of this is fixed in stone.

France has ordered 2.5 million civil servants off Microsoft Teams and Zoom, onto a homegrown French tool by 2027.

Germany built its own open source software, the very suite the ICC switched to after Microsoft locked it out.

Across Europe, spending on sovereign cloud, meaning systems that stay under home control, is forecast to jump 83 percent this year alone.

Britain is on the touchline, watching it happen.

Burnham, of all people, has the instincts for this. His whole reputation is built on the idea that you cannot just hand everything to the market and hope.

Digital sovereignty is that argument made real. It is the most natural fit imaginable for him.

But instincts are not policy. A good speech is not a procurement rule.

This is the real test

Burnham will be judged on the things everyone expects. The NHS. The cost of living. The boats in the Channel.

But the quiet inheritance, the one nobody put on a poster, may matter more than all of them.

We have built a state that runs on systems we do not own and cannot fully control.

Fixing that is not about hostility to America. Their technology is brilliant. That is rather the point.

It needs something simpler. And much harder.

Burnham says he wants to rewire Britain.

He could start with the wiring nobody can see.

Share Tech Meets Policy

Last week, the Open Rights Group warned the UK government about something most politicians would rather not think about. If Britain got into a serious disagreement with Washington tomorrow, over trade, over Greenland, over anything, the US could effectively shut down British public services by restricting access to the technology that runs them.

Subscribe now

That is not an exaggeration. It is a description of existing infrastructure.

The UK public sector spends £16.6 billion a year directly with technology suppliers. The overwhelming majority of that money flows to American companies. AWS and Microsoft each hold 30 to 40 per cent of the UK cloud infrastructure market, which means two companies headquartered in Seattle and Redmond, Washington, control the backbone of British government computing.

The numbers are staggering. On a single day in late 2023, AWS signed £894 million in contracts with the Home Office, HMRC, and DWP. That figure exceeded AWS’s entire previous decade of government cloud sales. Microsoft receives approximately £1.9 billion annually in government software licences. Google Cloud secured a £400 million sovereign cloud contract with the Ministry of Defence. The NHS is migrating 1.5 million staff onto Microsoft 365.

The Palantir problem

Then there is Palantir. Peter Thiel’s surveillance technology company now holds at least £670 million in UK government contracts across health, defence, policing, and nuclear weapons, with one MP telling Parliament that half of those contracts were never competitively tendered.

What makes the Palantir story particularly uncomfortable is the revolving door. An openDemocracy investigation found four MoD officials joined Palantir in 2025 alone. The most striking case was the MoD’s Director of Industrial Strategy, who helped write the Strategic Defence Review, left the department on 31 August, and joined Palantir nine days later. Three months after that, Palantir won a £240 million no bid contract.

I struggle to understand how any progressive government can look at that timeline and not see a problem. This is not about whether Palantir’s software works. It is about democratic accountability.

The kill switch is not theoretical

The argument against dependency becomes concrete when you look at what has already happened.

In February 2025, Trump sanctioned the ICC Prosecutor. Microsoft, complying with US law, moved to cut off his email access. The ICC was told that Microsoft would terminate services for the entire organisation if they did not comply.

The ICC replaced Microsoft with open source software developed by Germany’s Centre for Digital Sovereignty in October 2025.

Then there was Huawei. In January 2020, the UK approved Huawei for limited 5G access. Five months later, Washington imposed new sanctions, and the UK reversed course entirely, ordering complete removal by 2027 at an estimated cost of up to £2 billion. That decision was not made in London. It was made in Washington. The UK’s own security assessment was overridden by American export controls.

And the Anthropic standoff in February 2026 introduced a new precedent entirely. When the AI company refused to allow its models to be used for fully autonomous weapons, the Pentagon blacklisted it as a “supply chain risk,” a classification normally reserved for Chinese firms. A federal judge called the designation “Orwellian” and issued an injunction in March 2026. The Trump administration appealed days later. If the US government can do this to an American company for a policy disagreement, the implications for allied nations relying on American technology are obvious.

The legal architecture makes it worse

Here is what makes all of this technically dangerous, not just politically concerning. The US CLOUD Act, passed in 2018, allows American law enforcement to compel US technology companies to hand over data stored anywhere in the world. A warrant served on AWS in Virginia applies to data on AWS servers in London.

In June 2025, a Microsoft executive admitted under oath in France that the company cannot guarantee data sovereignty if legally required to comply with the CLOUD Act. That admission should have triggered emergency procurement reviews across Whitehall. Instead, the UK signed even more contracts.

Europe is building alternatives. Britain is not.

The Open Rights Group framed the issue in terms anyone can understand: just as relying on one country for energy would be irresponsible, so is overreliance on US companies for digital infrastructure. But the comparison with European allies makes it worse.

France ordered 2.5 million civil servants to abandon Microsoft Teams and Zoom in favour of a sovereign videoconferencing tool. Germany launched its own government cloud and developed the open source suite the ICC adopted. The EU issued a €180 million tender for sovereign cloud services and passed a resolution noting its reliance on non EU countries for over 80 per cent of digital products and infrastructure.

Britain, by contrast, has no coherent digital sovereignty strategy at all. The House of Commons Library confirmed in March 2026 that the UK Government has no overarching policy on digital sovereignty. UKCloud, Britain’s main indigenous government cloud provider, went bankrupt in 2022. There is no replacement. PM Starmer talks about being “an AI maker, not an AI taker.” The spending data tells a different story.

Gartner projects European sovereign cloud spending will surge 83 per cent in 2026, with Europe set to surpass North America as the largest sovereign cloud market by 2027. Britain is watching this happen from the sidelines.

This is not an anti American argument

I want to be clear about what this is and what it is not. This is not about hostility towards the United States or American companies. AWS, Microsoft, and Google build genuinely excellent technology. The question is whether any country should build its critical public infrastructure on systems whose legal and political obligations lie with a foreign government, especially one whose current leadership has shown it will weaponise technology access without hesitation.

The Conservatives created this dependency through a decade of outsourcing. Labour, despite its rhetoric about public sector reform, has deepened it. Neither party has treated digital sovereignty as a strategic priority. Neither has invested in domestic alternatives. Neither has asked the most basic question of resilient systems design: what is the fallback?

The technology exists. The alternatives are being built. France and Germany are proving it can be done. The only thing missing in Britain is the political will to start.

Share Tech Meets Policy

Last week, the United States government threatened to seize control of an AI company’s technology because it refused to remove safety restrictions.

Subscribe now

Defense Secretary Pete Hegseth gave Anthropic until 5:01 PM on Friday, February 27, 2026 to strip safety guardrails from its Claude chatbot or face consequences. Contract termination. National security blacklisting. Potential invocation of a Korean War era law to force compliance.

President Trump banned the company from all federal work hours after the deadline passed.

Then something even more disturbing emerged.

Less than 24 hours after Trump’s ban, the U.S. and Israel launched massive strikes against Iran, killing Supreme Leader Ayatollah Ali Khamenei. The Pentagon was still using Claude AI for intelligence assessments, target identification, and battle scenario simulations during the operation.

The same day Anthropic was blacklisted, OpenAI announced it had struck a Pentagon deal with functionally identical safety restrictions. And secured a record breaking 110 billion dollar funding round from Amazon, Nvidia, and SoftBank, valuing the company at 730 billion dollars.

This isn’t a story about one startup’s acceptable use policy.

It’s the first real test of whether anyone can say “no” to military AI. It’s about who decides how artificial intelligence goes to war. And it’s about what happens when a private company’s safety commitments collide with executive power in an age when algorithms can kill.

The February 2026 standoff between Anthropic and the Pentagon is the most consequential confrontation between Silicon Valley and the state since the NSA revelations of 2013.

And it raises questions no democracy has answered yet.

How we got here: Venezuela to Iran in eight weeks

The roots trace to July 2025.

The Pentagon awarded contracts worth up to 200 million dollars each to four frontier AI companies. Anthropic, OpenAI, Google DeepMind, and Elon Musk’s xAI. The goal was prototyping AI capabilities for national security.

Anthropic’s Claude became the first and only frontier AI model deployed on the military’s classified networks. It ran via Amazon’s top secret cloud and a partnership with Palantir.

The contract included two specific restrictions: no mass domestic surveillance of Americans, and no fully autonomous weapons.

The arrangement worked.

Until it didn’t.

On January 3, 2026, U.S. forces used Claude during Operation Absolute Resolve, capturing Venezuelan President Nicolás Maduro. When Anthropic inquired about Claude’s role, the Pentagon interpreted it as potential disapproval.

By mid February, Hegseth issued an ultimatum. All Department of Defence AI contracts must allow “any lawful use” within 180 days.

CEO Dario Amodei refused.

On Tuesday, February 24, Hegseth summoned Amodei to the Pentagon. The meeting included an unusually senior lineup: Deputy Secretary, two Under Secretaries, the chief spokesperson, and General Counsel. Hegseth gave Anthropic until 5:01 PM Friday to comply or face designation as a “supply chain risk to national security.”

Thursday, February 26. Anthropic rejected the final offer. Amodei stated the contract language “made virtually no progress on preventing Claude’s use for mass surveillance of Americans or in fully autonomous weapons”.

That evening, Under Secretary Emil Michael posted six times on X in two hours, calling Amodei “a liar” with “a God complex”.

But something unexpected happened. Over 100 Google employees sent a letter requesting similar limits on Gemini’s military use. Sam Altman told OpenAI employees internally that OpenAI shared Anthropic’s “red lines”.

Friday, February 27. The deadline day.

Trump posted on Truth Social: “The Leftwing nut jobs at Anthropic have made a DISASTROUS MISTAKE trying to STRONG-ARM the Department of War”. He banned Anthropic from all federal work. Hegseth designated them a supply chain risk, a label normally reserved for suspected foreign adversaries.

Hours later, OpenAI announced a Pentagon deal with functionally identical safety restrictions Anthropic had demanded. And secured 110 billion dollars in funding from Amazon, Nvidia, and SoftBank, valuing the company at 730 billion dollars.

The punishment, it appears, was not for the policy but for the defiance.

Then came Iran

Saturday, February 28. Less than 24 hours after Trump’s ban.

The U.S. and Israel launched Operation Epic Fury. Massive airstrikes across Iranian cities. Supreme Leader Ayatollah Ali Khamenei was killed.

The Pentagon was still using Claude.

U.S. Central Command used Claude for intelligence assessments, target identification, and combat scenario simulations. The AI processed thousands of hours of intercepted Persian communications, identified vulnerabilities in Iranian command chains, and generated strike scenarios.

Claude helped time the strikes to coincide with a rare Saturday morning meeting of senior Iranian officials in Tehran.

This was the first high level assassination operation in which AI played a central decision making role. Not auxiliary support. Core targeting and strategic planning.

The six month phase out Trump announced gave the Pentagon breathing room. Claude was so deeply integrated into Central Command’s systems that disabling it immediately would risk “combat control paralysis.”

What Anthropic actually wanted, and why it matters

Anthropic’s objections were narrow and technical.

The company supported Claude’s use for cyber defence, missile defence, intelligence analysis, and operational planning. It opposed two specific applications.

Fully autonomous weapons. Systems that identify, select, and engage targets without human intervention. Amodei’s objection was technical: frontier AI models are prone to hallucinations and unpredictable behaviour unsuitable for lethal targeting without human oversight.

Israel’s AI systems Lavender, Gospel, and Where’s Daddy reportedly identified 37,000 Palestinian civilians for targeting with an approximately 10 percent error rate. At least 14 drone models now have specifications suggesting autonomous targeting of individual humans.

Mass domestic surveillance. Using AI to aggregate location records, browsing history, social associations, and financial data from brokers into comprehensive portraits of citizens’ lives at scale.

The real question: who decides, and neither answer is good

The Anthropic standoff crystallises a genuine tension that defies easy progressive framing.

Unaccountable corporate power over defence decisions is undemocratic. A private CEO shouldn’t unilaterally determine national security policy.

But executive power to strip safety guardrails from AI is equally dangerous. Especially when deployed without legislation, without Congressional debate, without democratic accountability.

As Daniel Castro at the Information Technology and Innovation Foundation wrote: “The boundaries of military AI use should not be settled through ad hoc negotiations between a Cabinet secretary and a CEO. Democratic constraints on military AI belong in statute and doctrine, not in private contract negotiations”.

The United States has essentially no statutory framework governing AI in autonomous weapons or surveillance in a national security context. The Pentagon’s limits exist only in internal policy that can be changed without legislation.

Meanwhile, every other AI company quietly walked back safety commitments. OpenAI removed its prohibition on military use in January 2024. Google removed its pledge not to develop AI for weapons in February 2025. Meta opened its models to defence contractors in November 2024.

The UK is no better. The Strategic Defence Review 2025 committed over 4 billion pounds for autonomous systems. The Ministry of Defence’s dominant AI relationship is with Palantir, which hired four former MoD officials including the former Director of Policy who helped write the Strategic Defence Review. A textbook revolving door.

The EU AI Act explicitly exempts military AI from regulation.

Internationally, 164 countries voted in favour of UN resolutions on autonomous weapons in November 2025. Only six opposed: Belarus, Burundi, DPRK, Israel, Russia, and the United States.

This is how democracy loses control

The international precedent is the most dangerous dimension.

If the U.S. establishes that governments can force AI companies to remove safety guardrails, it legitimises China’s civil military fusion strategy. Russia’s approach to compelling tech company compliance. Any authoritarian regime’s demands on its domestic AI industry.

As Lawfare noted: “AI law enforcement tends to undermine democratic government, promote authoritarian drift, and entrench existing authoritarian regimes”.

The Anthropic standoff is not about one company’s acceptable use policy.

It is about whether democratic societies will create legitimate institutional frameworks for military AI before the technology outpaces every guardrail that currently exists.

The answer, so far, is that they have not.

And the Iran operation proved the urgency. When AI can help kill a nation’s supreme leader within hours of being supposedly banned, the question is no longer theoretical.

The Pentagon’s restrictions are administrative policy. The companies’ guardrails are voluntary. The executive’s power to override both remains legally untested.

Congress faces growing bipartisan pressure to act. The fact that Armed Services Committee leaders jointly urged restraint suggests even national security hawks recognise the current approach is unsustainable.

Retired Air Force General Jack Shanahan, who led Project Maven, captured the paradox: “Painting a bullseye on Anthropic garners spicy headlines, but everyone loses in the end”.

The Anthropic standoff is not really about one company’s acceptable use policy.

It is about whether democratic societies will create legitimate institutional frameworks for military AI before the technology outpaces every guardrail, corporate, executive, and international, that currently exists.

The answer, so far, is that they have not.

And the Iran operation proved the urgency. When AI can help kill a nation’s supreme leader within hours of being supposedly banned, the question is no longer theoretical.

The clock, unlike the Pentagon’s arbitrary 5:01 PM deadline, does not stop ticking.

Last week, Britain crossed a line that should alarm anyone who cares about technology, human rights, and the future of liberal democracy.

Subscribe now

Home Secretary Shabana Mahmood unveiled asylum reforms that read like a dystopian tech policy white paper. AI algorithms will assess whether refugee children are lying about their age. Biometric databases will track every movement. And even after the UK accepts you fled persecution, you’ll spend 20 years under constant threat of algorithmic deportation.

This isn’t just cruel immigration policy. It’s a pilot programme for how wealthy democracies will use AI and surveillance to create conditional citizenship based on economic productivity.

And because it’s being tested on refugees first, most people won’t notice until the infrastructure is permanent.

Facial Recognition Becomes Judge and Jury for Asylum Seekers

Here’s the most chilling element.

The UK is deploying facial age estimation AI to determine whether unaccompanied asylum seeking children are telling the truth about their age. Testing begins late 2025 with full integration planned for 2026.

Get classified as an adult by the algorithm? You’re sent to adult detention without safeguarding protections.

Trials in Australia already showed this technology exhibits racial bias. The AI was less accurate for certain ethnic groups.

But when your target population lacks political power, accuracy becomes optional.

Human Rights Watch didn’t mince words.

Meanwhile, Privacy International revealed something even more disturbing.

UK police facial recognition searches against immigration databases jumped from 16 in 2023 to 102 in 2024. This happened “without any safeguards or public debate.”

The surveillance infrastructure is already operational. This policy just makes it official.

When Refugee Protection Becomes an Economic Performance Review

But the AI surveillance is merely enforcement infrastructure for the policy’s real innovation.

Turning refugee protection into an economic performance evaluation.

Under the new system, recognised refugees spend 20 years under renewable 30 month permits before they can settle permanently.

Previously, refugees received 5 year initial grants automatically converting to indefinite leave to remain.

That’s the longest path to settlement in Europe. More than double Denmark’s 8 year model.

Reviews happen every 2.5 years with the burden on refugees to prove ongoing danger. One bad security assessment in your home country? Deportation, even after a decade of building a life in Britain.

Here’s where it gets economically perverse.

The government created an alternative “Protection Work and Study Route” allowing refugees who obtain employment to potentially earn settlement in 10 years instead of 20.

Read that carefully.

The bitter irony?

Asylum seekers are banned from working while claims are processed. Then once granted refugee status, they get less than 30 days to find housing.

The system creates the barriers that prevent economic integration, then punishes refugees for failing to integrate.

Government data shows over 50% of refugees remain on benefits 8 years after arrival.

Rather than asking why integration support fails, the policy assumes economic non productivity means you don’t deserve permanent protection.

The Denmark Model That Every International Body Condemned

The UK explicitly modelled this on Denmark’s restrictive approach, but went further in almost every dimension.

Denmark’s “paradigm shift” began in 2019 under Social Democratic Prime Minister Mette Frederiksen. She pursued a “zero asylum seekers” goal while framing restrictive migration control as protecting the welfare state.

The results?

Asylum claims fell to a 40 year low of around 1,500 to 2,100 annually, compared to over 21,000 in 2015.

Denmark claims a 95% removal rate of rejected applicants.

But here’s what the UK government didn’t mention in their glowing references.

Between 2019 and 2021, Denmark reclassified Damascus as “safe” and revoked protection for hundreds of Syrian refugees. Despite 11 of 12 Country of Origin experts condemning the decision.

Nearly 400 Syrians left Denmark in 2022, many fleeing to other EU states fearing status loss.

All international bodies disagreed with the “safe” designation. The UN, EU, UNHCR, and European Court of Human Rights.

Every single one.

The Council of Europe’s Anti Torture Committee described

Denmark’s Ellebaek Detention Centre as “not suitable for humans.” They cited excessive force, racism, and 15 day solitary confinement punishments. The report also noted restricted access to open air, sometimes only 30 minutes daily.

UNHCR issued 16 recommendations warning Denmark’s approach could trigger a “race to the bottom” if other countries follow.

European Council on Refugees and Exiles analysis noting Denmark is “not bound by minimum standards under the Common European Asylum System” due to opt outs. The European Council on Refugees and Exiles called it “rogue asylum policies.”

When Wealthy Democracies Deploy Technology to Abandon Asylum Obligations

This is where technology policy intersects with global inequality in troubling ways.

Only 39% of UK asylum claims in 2025 arrived by small boats. Another 37% arrived legally on visas first.

The “crisis” driving these reforms is partially manufactured.

More revealing: between June 2024 and June 2025, 58,000 asylum claims were refused but only 11,000 were actually removed.

That’s a 19% removal rate.

The government’s own research acknowledges “welfare policies and labour market access have little impact on migrant decision making.”

Yet the entire policy centres on making the UK less attractive through cruelty.

What’s really happening?

Wealthy democracies are using AI and surveillance infrastructure to offshore human rights obligations. Deploy algorithms to make it harder to claim asylum. Use economic productivity tests to sort “deserving” from “undeserving” refugees. Create such hostile conditions that people self deport.

It’s a high tech version of the same colonial logic.

Extract value from the Global South through economic systems and geopolitical instability, then deploy technology to prevent people fleeing the consequences from reaching your borders.

The Surveillance Infrastructure Tested on Refugees Becomes Everyone’s Future

Tommy Robinson endorsed these reforms.

Over 20 Labour MPs publicly dissented.

Nadia Whittome delivered the most devastating critique during the Parliamentary statement on 17 November.

“Is the Home Secretary proud that the government has sunk such that it is now being praised by Tommy Robinson?”

But here’s the broader lesson for anyone who cares about technology and democracy.

The mandatory biometric ID system being tested on refugees becomes the baseline for everyone by 2029.

The facial recognition databases built for immigration enforcement get repurposed for domestic policing. Searches already jumped 537% in two years, from 16 to 102 searches of immigration databases alone.

The algorithmic decision making models trained on asylum seekers become templates for welfare, housing, and employment systems.

Big Brother Watch warns

“We are alarmed by the recent escalation in the Labour government’s moves towards adopting a mandatory digital ID scheme. This crosses a dangerous redline past which our civil liberties can never fully recover.”

A parliamentary petition against mandatory digital ID gathered 2.9 million signatures.

The report also found 63% of the British public does not trust the government to keep their digital ID data secure.

Privacy rights erode at the margins first. What starts as “immigration control” becomes normalised surveillance infrastructure for everyone.

The Alternative That Actually Works But Requires Treating Refugees as Human

The Institute for Public Policy Research’s “Charting New Waters” report offers a comprehensive alternative.

Create safe legal routes through a Refugee Visa Pilot at embassies. Initially for Afghans where 97% of cases are granted asylum anyway. This would allow applicants to make initial assessments overseas and complete claims in the UK, reducing dangerous Channel crossings while maintaining territorial asylum obligations.

Fast fair decisions through adequate resources rather than AI shortcuts. End outsourcing of accommodation to private companies where costs soared from £17,000 per person in 2019/20 to £41,000 in 2023/24.

Genuine integration support rather than 20 year limbo.

Green MP Carla Denyer stated it simply.

“The only way to prevent people making dangerous crossings by small boats is to open safe and managed routes for people to claim asylum in the UK.”

But that would require treating refugee protection as a human rights obligation rather than an economic cost benefit analysis optimised by algorithms.

Conclusion: This Is How Liberal Democracy Dies Without Anyone Noticing

Labour’s asylum overhaul isn’t just about immigration.

It’s a blueprint for how wealthy democracies will use AI, biometrics, and economic productivity tests to create conditional citizenship.

Your right to remain depends on algorithmic assessments and market value.

Denmark pioneered temporary protection. The UK is adding AI surveillance and doubling the limbo period. Other European countries are already exploring similar models.

This is how democracies sleepwalk into authoritarianism.

Not through dramatic coups, but through incremental deployments of “efficiency enhancing” technologies on populations too vulnerable to resist.

Share Tech Meets Policy

That's how 85,000 digitally excluded refugees feel in modern Britain, according to Moses Seitler, founder and CEO of Screen Share UK. His organisation has become a lifeline in an increasingly digital world where being offline means being invisible.

While the government celebrated launching its £9.5 million Digital Inclusion Innovation Fund in August 2025 — the first comprehensive strategy in over a decade — the reality on the ground tells a different story, according to TechUK and GOV.UK. Screen Share UK received 6,000 referrals in the past 12 months but could only support 1,500 people. That's a 75% rejection rate for some of Britain's most vulnerable residents.

"The government has a massive blind spot when it comes to refugees and digital exclusion," Seitler explains. "The official statistics are completely inaccurate because they don't properly count asylum seekers in hotels, temporary accommodation, or those who are housing-insecure. We're dealing with people who are systematically excluded from the data collection itself."
(Interview with Moses Seitler, Aug 2025)

The numbers tell a stark tale of two Britains. While 8.5 million adults lack basic digital skills and 1.6 million remain completely offline, refugees and asylum seekers face compound exclusion that government policy barely acknowledges, according to the Good Things Foundation and Refugee Action. Despite recognising refugees' "significant and unique barriers" in consultation responses, the government’s Digital Inclusion Action Plan still fails to include them among its five priority demographic groups, as research by Rethinking Refuge has highlighted.

Northern England bears the brunt of digital apartheid

Geography is destiny when it comes to digital inclusion in Britain. Rouge Media reports that Northern Ireland leads with the highest exclusion rate at 11.8%, affecting 174,000 people. The North East follows at 11%, while London, despite having the lowest exclusion rate at 7%, still contains 347,000 digitally excluded residents.

Screen Share UK's data reveals crushing demand in these regions.

"We're seeing massive increases in referrals from London, the North East, North West, and Northern Ireland," Seitler notes. "People are desperate for basic connectivity and devices, but we can only offer seven-day fulfilment for our closest partners. Everyone else waits up to eight weeks."

The cost of digital exclusion hits hardest in these areas. According to the Good Things Foundation and GOV.UK, digitally excluded people pay up to 25% more for essentials like insurance, travel, and food. In regions already facing economic pressures, this digital poverty trap deepens existing inequalities. The mathematics are brutal: while the government invests £9.5 million nationally, digitally excluded households miss an estimated £560 in annual savings from online shopping and bill paying alone, according to TechUK.

The refugee digital apartheid hidden in plain sight

Mohammad, an Afghan refugee and father of two, described his family’s digital reality:

"Four of us had one phone to connect to the digital world. Job hunting was almost impossible without a digital device. All vacancies are advertised online, job applications are online."
(Refugee Action case study, 2022)

This is the reality for thousands invisible to official statistics. Asylum seekers receive just £7 per day while UK mobile data costs average £5.12 per gigabyte — among the world’s highest prices — according to Screen Share UK and the House of Commons Library. They are banned from working, cannot get broadband contracts without permanent addresses, and often rely on climbing to rooftops for mobile signal in poor-quality accommodation.

"I have to go to the roof of the building to get a connection," one asylum seeker told Refugee Action.

Meanwhile, the government's digital-first approach to Universal Credit and NHS services assumes everyone can navigate online systems — a luxury many refugees simply cannot afford.

Seitler's work shows the transformative power of digital inclusion when done right. After receiving a laptop through refugee support programmes, Mohammad completed multiple online courses and became a volunteer senior project manager helping other Afghan migrants.

E-waste mountains while people go without devices

Britain faces a digital contradiction bordering on the absurd. The country generates massive electronic waste, with weak reporting requirements allowing businesses to hide their disposal destinations. Meanwhile, thousands queue for refurbished devices costing £84 each, according to GOV.UK.

"We desperately need legislation requiring businesses to report e-waste destinations publicly," argues Seitler.

The current system allows corporations to dump perfectly functional equipment while charities like Screen Share UK scramble to meet overwhelming demand.

Windows 11 requirements have added artificial barriers, excluding functional computers older than seven years despite their capability to run essential software. This forced obsolescence drives perfectly good hardware into waste streams while digitally excluded people cannot afford “compliant” replacements.

The economics expose government priorities. Screen Share UK provides devices for £84 and year-long digital skills support for £150, yet receives 6,000 referrals while only managing 1,500 placements. Supporting all 8.5 million digitally excluded adults would cost £1.3 billion, but research from the Good Things Foundation and UK Parliament estimates this would generate a £23 billion economic benefit from improved digital skills for work alone.

International models show what’s possible

Hope exists.

Nordic countries demonstrate comprehensive digital inclusion through trained Digital Ambassadors, permanent community hubs, and user-centred design across government services. Singapore’s Smart Nation strategy supports over 150,000 senior citizens through culturally appropriate digital training, while India’s innovations have reached 800 million people through integrated digital identity, banking, and mobile connectivity.

These models prove digital inclusion works when “baked in” from design through delivery, rather than treated as an afterthought.

The solutions are hiding in plain sight

Britain needs systematic change, not charity. Digital inclusion must be treated as an infrastructure policy, not a social policy. That means mandatory Wi-Fi in all asylum accommodation, targeted data allowances for digitally excluded populations, and refugee-specific provisions in national strategies, as Refugee Action recommends.

"Digital inclusion should be baked in to support relationships from day one. When people have access to technology, they can improve their own lives, but only if we design systems that include rather than exclude."
(Moses Seitler, Aug 2025)

The path forward requires confronting uncomfortable truths. Current government investment represents roughly £1.12 per digitally excluded person — a sum that wouldn’t buy a coffee, let alone transform lives, according to UK Parliament. Meanwhile, Screen Share UK's waiting lists grow longer, regional disparities deepen, and refugees remain invisible to policymakers.

The choice before us

Britain stands at a crossroads: continue treating digital inclusion as charitable work for the “deserving poor,” or recognise connectivity as essential infrastructure for human dignity in the 21st century.

The evidence is overwhelming. The solutions exist. International models prove transformation is possible. The question isn’t whether Britain can afford comprehensive digital inclusion. It’s whether the country can afford not to act.

Every day of delay means more people climbing rooftops for signal, more students dropping out when courses move online, more job applications blocked by lack of devices, and more isolation in an increasingly connected world.

Subscribe now

Code and cloud are not neutral. As someone who studies distributed systems and cloud infrastructure, I see the exact building blocks we learn about being repurposed into tools of surveillance and control in Gaza through platforms like Microsoft Azure. This is not inevitable or technical junk. It is a political choice baked into architecture. When storage, streaming and real-time processing become instruments of domination, engineering decisions stop being abstract and start costing lives. We can design systems that protect people or systems that enable occupation.

How Peaceful Tech Powers War

Studying computer science means learning to see the world through patterns and systems. We deconstruct applications into their core components: data ingestion, processing, storage, and analysis. It is through this analytical lens that the reports from Gaza become not just a political tragedy, but a terrifying case study in the weaponisation of standardised cloud architecture. The very same distributed systems that I learn to build are being deployed, according to a joint investigation by The Guardian and +972 Magazine on 6 Aug 2025, to enact a digital hierarchy where Palestinian lives are reduced to mere data points in an Azure-powered surveillance system.

Their reporting indicates that Israel's Unit 8200 has constructed a massive, segregated Microsoft Azure environment designed to process intercepted Palestinian communications Al Jazeera reported 7 Aug 2025 and The Guardian reported 6 Aug 2025. The scale is almost incomprehensible: approximately 11,500 terabytes of intelligence data, equating to roughly 200 million hours of audio on Microsoft's servers in the Netherlands and Ireland according to a joint report by The Guardian and +972 Magazine published 6 Aug 2025. This is not passive storage; it is a fully operational, industrial-scale surveillance pipeline that transforms the most intimate moments of civilian life into actionable targeting data. The same Azure Speech services that power global customer service chatbots are transcribing private conversations, which determine life and death from the sky according to a joint report by The Guardian and +972 Magazine, 6 Aug 2025.

The Architecture of Oppression

The technical elegance of this system is undeniable. Microsoft's documentation describes capabilities that become terrifying when weaponised: the Microsoft Audio Stack processes up to 16 input channels with noise suppression and beamforming, Azure Speech converts conversations to searchable text, and petabyte-scale blob storage can retain everything indefinitely. This architecture follows a set of logic that captures, processes, categorises, and acts on predetermined notions of value and threat. Those categorisations shape who is seen and who is exposed. When infrastructure is built to sort people the way it sorts data, engineering stops being neutral and becomes policy with real human costs.

According to the investigation by +972 Magazine and Local Call, Unit 8200's alleged "Lavender" AI system reportedly marked 37,000 Palestinians as suspected militants with a 10% error rate—meaning 3,700 people were falsely identified. These aren't abstract algorithms; they're automation of extermination, powered by the same cloud infrastructure I'm learning to build.

The Complicity of Neutrality

Microsoft’s initial stance was a masterclass in corporate evasion. The company claimed it had "no evidence" its technologies were being used to harm people in Gaza (Microsoft On the Issues blog, 15 May 2025). This position became impossible to maintain after investigative reports detailed the extent of the Azure infrastructure supporting Unit 8200. The company was forced to commission an external review, a move that implicitly admitted its own systems were too opaque to monitor effectively (The Guardian, 9 Aug 2025).

This hypocrisy is stark. Microsoft’s own terms of service explicitly prohibit "broad or mass surveillance of civilians" (Microsoft On the Issues blog, 15 May 2025). Yet, internal documents reveal that executives, including CEO Satya Nadella, met with Unit 8200 commanders to discuss moving vast amounts of secret intelligence data into Azure (The Guardian and +972 Magazine, 6 Aug 2025). This is not a failure of policy, but a failure of principle. The very design of cloud computing, which isolates one customer's data from another for security, is being used as an ethical excuse for ignorance. Microsoft built a powerful tool, handed it to a military unit, and then claimed it could not see how it was being used. This is not a bug in the system; it is a deliberate and dangerous choice.

A Mandatory Human Rights Gate

The solution is not more after-the-fact reviews, but a fundamental change in process. To prevent cloud infrastructure from becoming an engine of oppression, its provision must be contingent on a mandatory, pre-deployment human rights assessment. For any contract with military or intelligence agencies in active conflict zones, cloud providers like Microsoft must require an independent Human Rights Impact Assessment before a single server is provisioned.

This is not a radical idea; it is basic due diligence. Microsoft’s own decision to finally commission an external review proves such assessments are feasible (Microsoft On the Issues blog update, 15 Aug 2025; The Guardian, 15 Aug 2025). The challenge is to make them a non-negotiable gatekeeper, not a public relations tactic deployed after atrocities are exposed. These assessments must be powerful and specific, evaluating not just the stated purpose of the technology but its potential for misuse, with a sharp focus on red flags like mass data ingestion and population-scale monitoring.

There Is No Neutral Server

We must abandon the comfortable lie that technology is neutral. A database is not a mere tool; it is a statement of priorities. The algorithms that power Azure are the same ones I study, and their potential for both good and harm is determined by the ethics of those who wield them. My learning journey has shown me that data is never just ones and zeros. In the wrong hands, it becomes a weapon.

This is not an abstract problem. The logic that reduces human beings to data points for analysis is the same logic that reduces people to stereotypes. It is a hierarchy of value that deems some lives less worthy of privacy, safety, and ultimately, existence. Microsoft and other tech giants must finally recognise that in building the infrastructure of surveillance, they are undeniably complicit in its consequences. Providing the tools is a choice. Profiting from their misuse is a choice.

Microsoft and other cloud providers must recognise that in providing the infrastructure of surveillance, they become complicit in its consequences. The time for neutral technicality has passed; we must choose which side of history our technology serves.

Share

Leave a comment

Follow on LinedIn

Photo by StefanCoders from Pixabay

Subscribe now

The numbers tell a devastating story of policy failure. Within hours of the UK's Online Safety Act taking effect on 25 July, Proton VPN reported a staggering 1,400% surge in British sign-ups. BankInfoSecurity Other providers saw similar explosions. NordVPN recorded 1,000% growth, BankInfoSecurity whilst five of the top ten free apps in Britain's App Store became VPN services. TechShout Meanwhile more than 500 000 Britons signed a parliamentary petition demanding repeal of the Act surpassing the 100 000 signature threshold needed for a debate. Petitions - UK Government and Parliment

This isn't simply a failure of current policy, but rather the predictable outcome of fundamentally flawed legislation inherited from the previous government. The Online Safety Act received Royal Assent in October 2023 Ofcom under Oliver Dowden's watch, Trilateral Research with Labour providing minimal opposition scrutiny. As LabourList observed, "the outgoing Conservative government left behind a mess" whilst "the Labour opposition provided no meaningful scrutiny." LabourList Now Peter Kyle and his team find themselves defending an unworkable surveillance system designed by their predecessors.

Technical failures expose surveillance overreach

The Act's age verification requirements have proven laughably inadequate. Users discovered that high-fidelity video game characters from Death Stranding can fool facial recognition systems designed to verify human ages. Thesmallbusinesscybersecurityguy The game's photo mode allows realistic expressions that bypass "liveness detection". The same exploit works with Baldur's Gate 3 and WWE 2K25 characters. thesmallbusinesscybersecurityguy

This technical farce demonstrates what cybersecurity experts warned: the Act creates "high-risk honeypots for hackers" whilst achieving nothing meaningful for child protection. Brian Higgins from Comparitech noted the "worrying" surprise among officials about VPN circumvention, a predictable outcome that Sarah Champion, Labour MP for Rotherham, specifically warned about during 2022 parliamentary debates. GB News

Champion's prescient concerns proved exactly right. She proposed New Clause 54 requiring government reporting on VPN circumvention effects, noting that "75% of children aged 16 and 17 used, or knew how to use, a VPN." The Tab Her amendment was rejected by Conservative ministers, and now Champion has posted on social media: "I did warn the last government this would happen." X - Sarah Champion's Tweet

Economic devastation hits communities hardest

LFGSS, a community with roughly 50 000 monthly users, is closing on 16 March 2025 along with some 300 other forums. Dee Kitchen cited the personal liability and legal exposure under the Online Safety Act as impossible burdens to sustain. The Guardian

These weren't commercial platforms but volunteer-run spaces fostering genuine community. Microcosm, hosting 300 small forums with 275,000 monthly users, closed because individual operators couldn't absorb £255,662 annual compliance costs. Techdirt

Meanwhile, Big Tech platforms that the Act supposedly targets have simply implemented basic age verification and continued operating. The legislation has achieved the opposite of its stated aims: consolidating power amongst tech giants whilst destroying grassroots digital communities.

Government rhetoric exposes authoritarian drift

Kyle's defence of the inherited Conservative policy has proved deeply troubling. When challenged about VPN circumvention, he told BBC Breakfast: "if you want to overturn the Online Safety Act you are on the side of predators. It is as simple as that." Reclaim The Net More shockingly, he compared critics to "people like Jimmy Savile," stating that "Nigel Farage is saying that he's on their side." Wikipedia

This inflammatory rhetoric deflects from policy substance whilst echoing authoritarian regimes that restrict VPNs. As Graeme Stewart from Check Point Software observed, "banning VPNs puts the UK in the company of China, Russia, and Iran." Fast Company The government hasn't banned VPNs yet but Kyle's warnings about monitoring VPN usage "very closely" suggest concerning future possibilities. The Register

International isolation demonstrates better alternatives

Whilst Britain pursues surveillance-based solutions, Europe has chosen privacy-preserving approaches. The EU's Digital Services Act emphasises risk assessments, transparency reporting, and privacy-by-design principles rather than mandatory content scanning. Their Better Internet for Kids Strategy focuses on education and empowerment alongside protection. European Commission

International expert consensus favours the European model. The Electronic Frontier Foundation describes Britain's approach as a "privacy disaster," Electronic Frontier Foundation whilst the UN Special Rapporteur has received formal complaints about freedom of expression threats. Wikipedia has launched High Court proceedings Wikipedia challenging its "Category 1" designation, warning of potential UK access restrictions.

The contrast couldn't starker: Europe protects privacy whilst achieving child safety goals through systemic approaches, whilst Britain's surveillance capitalism model fails technically whilst destroying community spaces.

A path forward for progressive governance

Labour shouldn't remain trapped defending Conservative policy failures. The evidence demands fundamental reconsideration: 500,000 petition signatures, Petitions 1,400% VPN usage surges, TechRadar community closures, and universal expert criticism prove the current approach unworkable.

A genuinely progressive response would adopt European privacy-by-design principles, focusing on education, mental health support, and platform transparency rather than mass surveillance. This means acknowledging that good governance requires admitting inherited policy mistakes, particularly when those policies undermine both digital rights and community spaces that progressive politics should champion.

The VPN explosion isn't just a technical failure; it's democratic resistance to poorly designed authoritarian overreach. PPC Land Labour has an opportunity to demonstrate that progressive governance means evidence-based policy reform, not doubling down on Conservative surveillance capitalism that benefits tech giants whilst harming ordinary communities.

The choice is clear: continue defending the indefensible or forge a genuinely progressive path that protects children without sacrificing democracy.

Share

Leave a comment

Follow on LinkedIn

Subscribe now

Starlink isn't just providing internet, it's creating digital dependence. With over 7,600 satellites forming the world's largest constellation, Elon Musk's system has become the digital infrastructure equivalent of colonial railways: built supposedly for progress, but designed for extraction and control. The technical setup behind satellite internet creates never-seen-before ways for watching and controlling that ground-based networks never could.

The most worrying aspect isn't the technology itself, it's how few people understand why satellite internet works completely differently from the cables and towers we're used to, and why this difference matters for global power.

The invisible architecture of orbital control

Unlike ground-based internet infrastructure that governments can regulate through physical access to cables and data centres, satellite internet operates through a technical setup that puts control in the hands of one company in ways that would make authoritarian regimes jealous.

Low Earth Orbit (LEO) satellites like Starlink operate at just 550 kilometres altitude. That's close enough to provide 20-45 millisecond delays that match fibre optic cables, but far enough to remain beyond any single nation's control. This creates a technical sweet spot that traditional geostationary satellites at 35,786 kilometres could never achieve, with their 550+ millisecond delays that made real-time applications impossible.

But here's what makes LEO constellations uniquely powerful: every bit of data must flow through ground stations controlled by the satellite operator. Unlike the internet's distributed setup where traffic can route through multiple providers and countries, satellite internet creates natural bottlenecks. Starlink operates these gateway stations across 40+ US states, meaning all global traffic eventually flows through American soil before reaching the broader internet.

The satellite-to-satellite laser links connecting Starlink's mesh network at 200 Gbps speeds make this control even stronger. While marketed as reducing dependence on ground infrastructure, these links actually enable remote geographic control through software-based boundaries. The system divides Earth into hexagonal cells approximately 20 kilometres across, allowing GPS-precise service boundaries that can be changed instantly from SpaceX headquarters.

Digital colonialism through pricing and punishment

The economic extraction model reveals satellite internet's colonial character. Starlink charges Americans $120 monthly while offering Kenyans the same service for $10, not from kindness, but to rapidly capture markets and create dependencies before competitors arrive.

This pricing strategy mirrors historical colonial approaches: initial cheap pricing to destroy local competitors, followed by extraction once alternatives are removed. Traditional satellite providers like Viasat have lost over 50% of subscribers since Starlink's launch, while local ISPs across Africa struggle to compete with Starlink's suddenly affordable pricing that undercuts their infrastructure investments.

The punishment mechanisms reveal the system's true nature. When Iran experienced internet blackouts following Israeli strikes in June 2025, Musk activated Starlink with a simple post: "The beams are on." Iran's parliament responded by banning the service entirely, threatening users with flogging and imprisonment, recognising that allowing American-controlled internet infrastructure represented a direct threat to their control over their own affairs.

Brazil's Supreme Court discovered this control setup firsthand when attempting to enforce its X (Twitter) ban. Initially Starlink refused compliance, demonstrating how satellite internet can bypass national authority. Only when faced with asset freezes and $920,000 daily fines did SpaceX give in, revealing that economic pressure on the controlling entity remains the only workable enforcement mechanism.

The technical impossibility of resistance

Building competing LEO constellations requires overcoming barriers that make resistance nearly impossible for most nations. The capital requirements alone, $10+ billion for full setup, exceed the GDP of 60+ countries. But money represents just the beginning.

SpaceX's vertical integration creates growing advantages. While competitors like Amazon's Kuiper project must coordinate across multiple launch providers, SpaceX launches its own satellites using reusable Falcon 9 rockets, achieving cost advantages that external providers cannot match. Starlink produces over two satellites daily at peak efficiency, a manufacturing rate that requires consumer electronics-scale production capabilities.

The technical complexity adds to these barriers. Each satellite requires advanced directional antennas (which can focus signals precisely like a spotlight), satellite-to-satellite laser communications, automatic collision avoidance systems, and sophisticated beam-steering technology. The system performs over 250 collision avoidance moves daily while managing dynamic routing across thousands of moving nodes, technical challenges that traditional satellite operators never faced.

Spectrum allocation creates additional bottlenecks. LEO orbital slots operate on "use it or lose it" ITU regulations, requiring 50% constellation setup within specific timeframes. With prime orbital real estate increasingly claimed, later entrants face inferior positioning and higher interference risks.

Surveillance architecture that governments envy

The monitoring capabilities built into satellite internet setup go beyond anything ground-based networks provide. Every user terminal sends out precise GPS coordinates continuously, a technical requirement for maintaining satellite connectivity that creates complete location tracking.

Unlike ground-based ISPs that can be bypassed through VPNs or alternative routing, satellite internet provides operators with complete visibility into user behaviour. Research shows that satellite communications can be intercepted using £400 worth of commercially available equipment, while the centralised ground station setup enables complete traffic analysis and deep packet inspection (which means reading the contents of internet communications).

The Ukraine conflict provided the clearest demonstration of these control mechanisms. SpaceX put geographic boundaries around Crimea to prevent Ukrainian naval operations against Russian assets, while restricting military use of the system for direct offensive purposes. Critical military communications depended on decisions made by a private individual rather than government policy, a precedent that should terrify anyone concerned about democratic accountability.

Breaking the orbital chains

Despite these strong barriers, technical solutions exist for developing countries willing to pursue digital control over their own affairs. Open-source satellite platforms like the Open Source Satellite Programme offer 50kg satellites for under £1 million, dramatically cheaper than traditional approaches. PocketQube technology reduces launch costs to just £25,000 per tiny satellite, making regional constellations economically workable.

Regional cooperation represents the most promising approach. The African Continental Free Trade Area and ASEAN frameworks already include telecommunications cooperation provisions that could support joint satellite development. The World Bank has identified a £100 billion investment need for African connectivity that could be met through coordinated regional satellite infrastructure rather than dependence on American systems.

Ground-based alternatives deserve equal attention. High Altitude Platform Stations operating at 20,000 metres altitude can provide satellite-like coverage with ground-based control and significantly lower costs. Fixed wireless ISPs using 5G technology offer high-speed connectivity without satellite dependencies, while hybrid setups can optimise between ground-based and space-based solutions based on local conditions.

The technical path to independence requires both immediate and long-term commitments. Countries must establish regulatory frameworks that promote competition while supporting local capability development. Open-source satellite ecosystems, shared ground station networks, and technology transfer programmes can build local expertise while reducing individual country costs.

The choice facing the Global South isn't between connectivity and isolation, it's between digital control over their own affairs and digital colonialism. Satellite internet's technical setup makes resistance difficult but not impossible. The question is whether developing nations will recognise the threat in time to build alternatives, or whether they'll wake up to find their digital infrastructure owned by a handful of American billionaires who can switch off their connectivity with a tweet.

The beams may be on now, but who controls the switch matters more than the signal itself.

Share

Leave a comment

Follow on LinkedIn

Photo by prez_12 from WallpaperCave

Subscribe now

On 14 February 2025, two Serbian investigative journalists received what appeared to be harmless Viber messages that would demonstrate just how close surveillance reaches into our daily lives. The messages contained Pegasus spyware infection links, crafted with intimate knowledge of their ongoing corruption investigations. Both journalists worked for the Balkan Investigative Reporting Network, exposing government misconduct.

The attack wasn't sophisticated in its social engineering. The Serbian phone number was registered to state-owned Telekom Srbija, revealing the brazen confidence of operators who no longer fear discovery. Amnesty International's technical analysis confirmed what civil society already suspected: governments are weaponising the very communication networks citizens rely upon for democratic discourse.

This incident illuminates a broader reality. Surveillance technology has evolved beyond the paranoid fantasies of 20th-century authoritarianism into something far more extensive and technically sophisticated. We are witnessing the emergence of what scholars call "digital authoritarianism" - a governance model that uses technology not just to monitor dissent, but to reshape the very possibilities of democratic participation.

When Smartphones Become Snitches

Your phone is no longer just a communication device. It's a potential surveillance platform that governments can activate remotely, without your knowledge or consent. NSO Group's Pegasus spyware represents the apex of this technological capability, and its latest iterations reveal how completely personal privacy has been compromised.

The 2022 PWNYOURHOME exploit demonstrates the terrifying sophistication of modern surveillance. This zero-click attack targets iOS devices through a two-phase process: first exploiting Apple's HomeKit daemon, then manipulating iMessage vulnerabilities. Victims need to do nothing - no malicious link clicking, no suspicious app downloads. Simply owning an iPhone running iOS 15-16 makes you vulnerable.

The technical elegance is as disturbing as it is impressive. Pegasus bypasses Apple's Pointer Authentication Codes by repurposing signed pointers in the iOS shared cache, executing malicious code through the very security mechanisms designed to protect users. It can read messages, access cameras and microphones, harvest passwords, and track location - all while scrubbing iPhone log files to evade detection.

NSO Group's business model reveals the manipulation of surveillance. Government clients pay between $3 million and $30 million for Pegasus access, with Saudi Arabia alone spending $55 million and Mexico $61 million over multiple years. This type of spying is not boutique - it's industrialised surveillance available to any government with sufficient budget.

Despite NSO's recent legal troubles including a $168 million judgment from WhatsApp, the company continues operating with 350-380 employees. There are no signs that the technical warfare will stop, with new zero-click exploits emerging regularly to defeat Apple's latest security measures.

The Infrastructure of Control Is Already Built

The surveillance system is not being built. It's operational and expanding. China's City Brain systems process over 20 million images daily through AI command centres that can identify "abnormal crowd behaviour" and track individuals across multiple locations. This isn't science fiction; it's the reality of urban governance in 2025.

Russia's approach differs but proves equally effective. The deployment of 6,000+ TSPU devices across Russian networks enables Deep Packet Inspection technology that can block 90% of civilian VPNs, with plans to achieve a 96% block rate by 2030. Unlike China's centralised model, Russia employs "decentralised deployment, centralised control," allowing Roskomnadzor to block websites separately across multiple networks.

Iran's Noor Plan represents perhaps the most chilling example of technology-enabled social control. Launched in April 2024 following the Mahsa Amini protests, the system has already resulted in at least 618 arrests for hijab violations. The technical implementation combines aerial drone surveillance, facial recognition software, and mobile app-based public reporting through the government-developed Nazer platform.

The system's integration with existing infrastructure demonstrates how surveillance capabilities can be rapidly scaled. Traffic cameras analyse vehicles for violations, triggering real-time text messages to owners. University entrance gates monitor female students through facial recognition. The proposed Hijab and Chastity Bill would require all private sector entities to provide CCTV footage to law enforcement, creating a comprehensive surveillance network that encompasses every aspect of daily life.

Algorithms Are Becoming the New Censors

The automation of censorship represents a qualitative shift in how information control operates. China's DeepSeek AI system alters or suppresses 85% of outputs on sensitive topics, trained on 133,000+ censorship examples to align with Chinese Communist Party preferences. This isn't human moderators making contextual judgments - it's algorithmic control at unprecedented scale and sophistication.

Twenty-one countries now require or reward platforms to deploy machine learning for content removal, while forty-one countries have blocked websites containing protected political, social, or religious speech - a record high. The European Union's Digital Services Act requires automated content moderation, effectively privatising censorship by making platforms liable for user content.

The technical capabilities are expanding rapidly. Natural Language Processing analyses text while computer vision processes images and videos. These systems can process billions of posts daily, applying machine learning pattern recognition to identify "harmful" content in real-time. The integration with surveillance infrastructure means social media monitoring can trigger visa revocations, immigration enforcement, or criminal investigations.

This Isn't Just a Dictator Problem

Democratic governments are rapidly adopting authoritarian surveillance methods, often justified through counter-terrorism or immigration enforcement. The US "Catch and Revoke" program has resulted in 300-600 visa revocations as of mid-2025, using AI to monitor foreign nationals' social media for perceived support of designated terrorist organisations.

Palantir Technologies exemplifies the public-private surveillance complex. The company's nearly $1 billion Navy contract and $178 million Army contract demonstrate how surveillance capabilities are being integrated across government agencies. Palantir's Foundry platform creates master databases linking tax returns, employment information, immigration status, family composition, social media activity, and biometric data.

The global nature of this market reveals how Western companies enable authoritarian repression worldwide. NSO Group's Pegasus has been sold to governments in 45+ countries. Swedish company Teliasonera installed "black boxes" enabling unrestricted surveillance in Belarus, Uzbekistan, Azerbaijan, and other repressive states. Canadian/US company Sandvine sold internet surveillance products to Belarus, Egypt, and Uzbekistan before announcing withdrawal from 56 "non-democratic" countries by end of 2025.

This represents more than technology transfer. It's the creation of a global surveillance infrastructure where democratic and authoritarian governance models converge around shared technological capabilities and operational practices.

Can Tech Be the Cure, Not the Disease?

Resistance is possible, and it's happening through both technical innovation and policy intervention. Signal has grown to 70-100 million monthly active users, demonstrating sustained demand for private communication despite government pressure. The messaging app's quantum-resistant encryption and zero-knowledge architecture provide genuine protection against surveillance, even attracting CIA employees who use Signal by default with standard onboarding training.

The Tor network handles 2 million daily direct connections through 6,000+ volunteer relays, successfully resisting blocking attempts through bridge relays and the Snowflake project. This represents more than technical resilience. It's a decentralised infrastructure for digital resistance that operates beyond any single government's control.

The EU AI Act, which entered force in August 2024, represents the most significant regulatory response to surveillance technology. Real-time biometric identification in public spaces, social scoring systems, and untargeted facial recognition database creation are now prohibited across the European Union. While implementation remains challenging, the legislation establishes crucial precedents for limiting AI-enabled surveillance.

Civil society resistance is evolving beyond individual privacy tools towards systematic challenges to surveillance infrastructure. The Surveillance Technology Oversight Project has organised 100+ organisations for oversight legislation, while Privacy International documents the public-private partnerships that enable digital authoritarianism.

The future depends on whether technical countermeasures can outpace surveillance capabilities, and whether democratic societies can develop the political will to limit their own governments' surveillance powers. The tools of resistance exist, but they require widespread adoption, sustained funding, and continued innovation. Democracy's survival in the digital age depends on citizens actively choosing privacy over convenience, and governments choosing transparency over control.

Share

Leave a comment

Follow on LinkedIn

Photo by ClickerHappy from Pixabay

Subscribe now

When Zizheng Pan finished his internship at NVIDIA in Silicon Valley during summer 2023, the tech giant was ready to offer him a full-time position. The 32-year-old AI researcher had all the credentials: a PhD from Monash University, proven research skills, and the backing of senior NVIDIA scientists. But Pan chose DeepSeek without much hesitation, returning to China to join what was then a little-known startup with just three people in its multimodal team. Rest of World

Eighteen months later, DeepSeek surpassed ChatGPT in App Store downloads, sending shockwaves through Silicon Valley. China Academy Pan posted on X: "This moment is absolutely phenomenal to me." Rest of World His mentor at NVIDIA, Zhiding Yu, reflected on the broader implications: "Many of our best talents come from China, and these talents don't have to succeed only in American companies. Instead, we learn a lot from them." Rest of World

Pan's story illustrates a seismic shift in global AI talent flows. While the United States and United Kingdom tighten immigration controls, China is systematically attracting the world's brightest AI minds. The consequences extend far beyond individual career choices, reshaping the global balance of AI power.

The Trillion Dollar Talent Shortage

Artificial intelligence represents the defining technology of our era, with McKinsey estimating its economic potential at $2.6 trillion to $4.4 trillion annually. McKinsey & Company Yet a critical bottleneck emerges: talent. The AI job market exploded with 59% growth in 2024, creating 16,591 new positions in the United States alone. Axios AI engineers now command median salaries of $156,998, with top performers earning over $300,000 annually. Veritone

This premium reflects genuine scarcity. The United States faces a projected shortage of 700,000 AI positions by 2027, CSET while Germany expects 70% of AI jobs to remain unfilled. Staffing Industry The skills gap is so severe that 63% of IT leaders identify AI and machine learning as their biggest talent shortage. DataCamp

Why are these workers so valuable? AI development requires rare combinations of advanced mathematics, computer science expertise, and domain knowledge. Training a single AI researcher takes years, and the pool of truly elite talent numbers in the thousands globally. These individuals design the algorithms that power everything from smartphone cameras to autonomous vehicles to medical diagnostics.

America and Britain’s AI Lockout

Both the United States and United Kingdom have erected significant barriers to AI talent immigration. The U.S. H-1B visa program, crucial for tech workers, approved just 14.6% of applications in 2024, down from 46.1% in 2021. Brookings Chinese nationals face particularly harsh treatment, with processing delays exceeding 2.5 years and systematic visa revocations targeting students in AI-related fields. Rest of World

The human cost is striking. Kai Chen, a Canadian AI researcher crucial to OpenAI's GPT-4.5 development, was denied a green card in early 2025 despite 12 years in the United States. Newsweek Forced to work remotely from Vancouver, she wrote: "Really sucks to get denied after waiting for so long and unable to return home." Newsweek

The United Kingdom has implemented equally restrictive policies. Post-Brexit immigration changes raised salary thresholds for skilled workers from £38,700 to £41,700, while extending the path to permanent residence from five to ten years. NHS Employers Immigration costs have surged 126% since 2019, making the UK the second most expensive destination globally for skilled workers. TechUK

These policies reflect political pressures rather than economic logic. The Trump administration's return brought immediate changes, with over 1,000 international students having visas revoked in 2025, including AI doctoral candidates. TechCrunch The government now uses AI systems to screen visa holders, creating a cruel irony where artificial intelligence blocks its own creators from entering the country. TechCrunch

China’s Calculated Bet on AI Brains

While the West builds walls, China constructs bridges. The country's Thousand Talents Program has attracted over 7,000 elite researchers since 2008, offering million-yuan bonuses and substantial research funding. Foreign Policy Current talent programs span over 200 initiatives across central, provincial, and city levels, creating a comprehensive ecosystem for AI development.

China's educational infrastructure provides massive scale. Since 2018, 535 universities have established AI undergraduate majors, making it the country's most popular new field. Asia Times The nation now produces 47% of the world's top AI researchers, up from 29% in 2019, while the United States has dropped to 18%. Brookings

The results are transformative. Chinese institutions published 23,695 AI papers in 2024 compared to 6,378 from the United States. Science China accounts for 61.1% of global AI patent origins and maintains 30,000 AI researchers versus 10,000 in America. Brookings The research pipeline continues strengthening, with Chinese AI researchers notably younger than their Western counterparts. Science

DeepSeek exemplifies this success. The company recruited primarily from elite Chinese universities like Tsinghua and Peking University, offering competitive salaries and academic-style collaboration. Rest of World Its breakthrough models emerged from a team of 31 top contributors averaging over 1,500 citations each. Rest of World Unlike Silicon Valley's visa uncertainties, these researchers faced no immigration barriers.

The Market Impact of Losing AI Minds

The economic stakes are enormous. AI-exposed sectors experience five times higher labour productivity growth, while companies with AI capabilities show significant valuation premiums. PwC McKinsey reports that 40% of their projects now involve AI, while Accenture invested $3 billion to double their AI workforce to 80,000 people. Medium

Immigration restrictions impose substantial costs. Foreign nationals earn 59% of AI-relevant PhDs in the United States, yet 60% report immigration difficulties. Over half of non-citizen AI PhDs are leaving America, with immigration issues cited as the primary reason. Brookings The brain drain accelerated dramatically: scientists of Chinese descent are leaving the United States 75% faster since 2018. Inside Higher Ed

Microsoft's recent dilemma illustrates the challenges. The company asked 700-800 Chinese employees in AI and cloud computing to relocate from China to other countries by July 2024. Fewer than half accepted, citing concerns about pay cuts, higher living costs, and visa uncertainty. Rest of World One Beijing-based engineer explained: "Emigration is not that appealing to many Microsoft people in China. If you deduct taxes, every place except Seattle may come with a pay cut compared to Beijing." Rest of World

The Path Forward

As artificial intelligence reshapes global competition, talent becomes the ultimate strategic resource. Countries that attract and retain the world's brightest AI minds will define the next century's technological landscape. Those who erect barriers will find themselves increasingly isolated from the innovation that drives economic growth.

Several policy solutions could address these challenges. The United States should expand H-1B visa caps or create dedicated categories for AI talent, while streamlining processing for exceptional ability applications. Kodem Law The United Kingdom could expand its High Potential Individual visa list beyond 20 American universities to include leading global AI institutions. CSET

Both countries need comprehensive immigration reform prioritising economic competitiveness over political rhetoric. Canada and Australia have gained ground by offering clearer pathways to permanent residence, while countries like Germany and the Netherlands attract talent through streamlined processes and cultural openness. CSET

The private sector must also adapt. Companies should diversify visa strategies, invest in premium processing, and consider international offices in countries with favourable policies. Immi-USA Academic institutions need stronger partnerships with industry to expand the talent pipeline.

While politicians debate border security, the real battle for national competitiveness plays out in university labs and corporate research centres. In this race, border politics may prove as crucial as any algorithm.

Share

Leave a comment

Follow on LinkedIn