How should Competition and Consumer Protection rules evolve in the age of

Artificial Intelligence?

Abstract

This essay explores the need for competition and consumer protection rules to evolve

in the age of rapid and unprecedented AI growth.

Section 1 establishes the context of the novel challenges posed by AI, highlighting an

urgency for relevant regulations to be reformed. It emphasises the delicate balancing

act required to ensure that these changes, while effectively addressing unorthodox

challenges, do not unduly hinder innovation. Topics of algorithmic collusion, merger

and acquisition (M&A) activities, self-preferencing and price discrimination are key

areas we have identified for further discussion to formulate improved legal

frameworks.

Section 2 explores how competition rules should evolve to better cope with the

demands of the AI age. This includes refinements to better regulate algorithmic

collusion and M&A enabled by AI – actions that often go unnoticed because

technology has allowed such behaviour to be carried out in subtle and inconspicuous

manners. In particular, we urge for the broadening of definitions of certain legal terms,

as well as the importance of establishing regulatory bodies specifically targeted

towards anticompetitive AI behaviour.

1
Section 3 analyses the role of consumer protection laws in safeguarding public

interest. Practices such as self-preferencing and personalised pricing have long been

present, and are only going to evolve further and quicker, and possibly in more

harmful ways with the aid of AI and algorithms. This can exacerbate existing issues of

information asymmetry. Additionally, data concerns regarding consumers have been

identified as an overlapping issue across both sectors.

As such, Section 3 seeks to suggest redresses and modifications to current consumer

protection laws to better capture the swift-changing market dynamics empowered by

AI. This includes the redefinition of relevant terms, as well as the introduction of

explicit provisions for data privacy in protection laws.

(286 words)

2
1. Introduction

The transformational growth of artificial intelligence (“AI”) has demanded adjustments to

antitrust laws to stay relevant. While Singapore’s adherence to free market economics

and principles has aided in its quest to foster innovation, even market purists recognise

that the ideals of a free market have to be tempered to achieve broader societal goals.

Hence, dealing with the idiosyncratic challenges posed by AI requires a deliberate

calibration of rules to safeguard the interests of firms and consumers without

unnecessarily stifling development.

Cognisant of the need to identify the prevalent challenges and potential benefits to be

reaped in the age of AI, this paper seeks to identify gaps within the current antitrust

structure. Namely, examining algorithmic collusion, merger and acquisition activities,

self-preferencing and price discrimination.

3
2. Competition Concerns in the AI Economy

2.1. Algorithmic Collusion

With the growing integration of AI in the economic sphere, market dynamics and

competition have inadvertently been fundamentally altered. The usage of AI in markets

introduces potential risks of anti-competitive behaviour from an antitrust standpoint,

specifically, algorithmic collusion (Bird & Bird, 2024). Facilitating this risk is the increase

in transparency, swiftness of business decisions, and the instantaneous manner in

which firms can respond to rivals’ actions (Hawkes, 2021). Spurred by the rapid

dissemination of information and the real-time nature of AI-driven decision making

processes, they have compounded opportunities for both explicit and tacit collusion to

be effortlessly sustained (University of Oslo, 2022).

There are four such scenarios where algorithmic collusion may arise (Ezrachi and

Stucke, 2017):

Messenger AI executes human directives and enables established

collusion through conventional methods stemming from

human interaction.

Hub and Spoke Rivals form vertical arrangements with an AI software

developer (the "hub"), aiding in the coordination of anti-

competitive horizontal agreements among counterparties

("spokes"). This arrangement arises from the algorithm

provided by the developer, rather than direct communication

4
 among competitors.

Predictable Agent Firms unilaterally employ their individual pricing algorithms.

Yet, these algorithms respond to each other’s prices,

performing as “predictable agents”. This raises the risks of

conscious parallelism or tacit collusion.

Digital Eye The threat of highly advanced self-learning AI independently

making profit-maximising decisions. Though human

interference is absent from this process, it still results in

collusion.

Such scenarios are already occurring and have been subjected to crackdowns by anti-

competition agencies. For example, in 2023, the Italian Competition Authority launched

an ex-officio investigation regarding the usage of pricing algorithms in passenger air

transport (Bird & Bird, 2023). Thus, there has been a metamorphosis of anti-competitive

conduct from the “smoke-filled rooms” of the past, to a dynamic environment in which

complex algorithms are able to manipulate business strategies without the need for

firms to enter any overt agreement (University of Oslo, 2022).

To better address these novel challenges, current antitrust laws have to evolve in two

ways:

5
 a. Broadening the term “collusion” under competition law due to the potential of AI

enabling competition restricting practices through lawful means

Given the blurring delineation between acts of unlawful explicit collusions and the

nuanced legality1 surrounding practices that form tacit collusion, there has been

an increasing risk of AI augmenting the risk of tacit collusion in non-oligopolistic

markets (Ibid.).

For instance, Section 34 of Singapore’s Competition Act necessitates the

evidence of “agreements, decisions and concerted practices” to classify as an act

of violation. Such a definition does not encapsulate all situations that could

potentially lead to anti-competitive effects. This trend is particularly pronounced

in industries where AI is heavily utilised for pricing strategies and market analysis

(Calvano, E., Calzolari, G., Denicolò, V., & Pastorello, S., 2020). Profit-

maximising algorithms can inadvertently lead to parallel pricing behaviour among

competitors (HBR, 2021), replicating the effects of collusion without any explicit

communication between firms.

Simply put, the legal loophole created runs the risk of not being addressed by

current antitrust laws. Additionally, the opacity of AI decision-making processes

can render it challenging for regulators to discern between intentional collusion

1
Despite its socially undesirable nature (resulting in anti-competitive effects that can manifest through
higher prices), tacit collusion is not automatically illegal. It can occur organically in a market due to
economic conditions and market dynamics.

6
 and algorithmic convergence (Skrine, 2024), further complicating efforts to

maintain fair competition.

To remedy this problem, there is an urgent need to redefine the term collusion.

Though the deliberately broad angle in which this term is currently construed

enables the prohibition of explicit coordination, it fails to address subtler forms of

tacit understanding or algorithmic collusion. Expanding the scope of its normative

definition to include implicit coordination and algorithm-driven market behaviour

would better enable regulators to have the necessary tools to address anti-

competitive conduct.

b. Adjust existing concepts of antitrust liability due to the autonomous nature of AI

(Molski, 2024)

As technology inches closer to the development of independent AI systems, this

calls into question the matter of liability surrounding the undertaking of AI

attributes. The reduction of human dependence by AI-driven processes

necessitates a shift in a legal system that is centred around the basis of human

agency and accountability.

Taking into account the unique characteristics of AI systems (such as their

autonomy, decision-making algorithms, and capacity for learning and

adaptation), legal frameworks should be redefined. This ensures that legal

7
 standards remain relevant and effective in governing the actions and

consequences of increasingly autonomous technologies.

2.2. Mergers and Acquisitions (M&A)

The usage of M&A in the emerging AI sector is a potential weapon in diminishing

competition within markets (Bird & Bird, 2024). Horizontal transactions in which

established firms acquire budding competitors could result in “killer acquisitions2” that

threatens competition (MayerBrown, 2024). With rising numbers of cloud service

providers developing proprietary AI models and making substantial investments3, there

are mounting concerns over the entrenchment of market power and the potential of

excluding downstream competitors (Skadden, 2024).

Foreclosing competition through the usage of vertical transactions to cut off other firms’

access to key inputs (Bird & Bird, 2024) is also a potential harm. Given the foundational

role of data in the creation of Generative AI (GenAI) (Weforum, 2023), such acts would

raise the barriers to entry, hindering the establishment of an equalised playing field for

entrants to compete effectively.

Since traditional methods of evaluating mergers rely on static market definitions, it fails

to encompass the dynamic forces shaping long-term competition in emerging AI

2
Firms acquire nascent competitors only to discontinue the target's innovation projects, thereby pre-
empting the emergence of future competition (OECD, 2021).
3
Google, Microsoft and Amazon boosted investment by half over three years to a combined quarterly
total of $32bn (Financial Times, 2024).

8
industries (Frontier Economics, 2023). As such, conventional instruments lack the

necessary nimbleness to assess mergers of firms in fast-paced markets.

A suggestion could be the introduction of ex-ante regulation tailored to the AI sector to

proactively address concerns. Through implementing pre-approval requirements,

thresholds4 can be set for the notification of M&A. Acquisitions should also be examined

to stop smaller buyouts that may aid firms in accumulating market power. This could

also entail establishing specialised regulatory bodies with technical expertise in AI

technologies. Market dynamics can be better monitored. Hence, mitigating the risk of

market consolidation through preemptively addressing potential threats to competition.

4
Based on market share, revenue or transaction value

9
3. Consumer Concerns in the Age of AI

In the AI age, companies’ control over data may create barriers to entry that prevent fair

competition from fully flourishing. While technological advancements have made

consumer spending and information gathering more convenient, they can also harm

consumers by facilitating unprecedented modes of anti-competitive conduct (Mintz,

2023).

In this section, we identify two areas that are prominent causes of concern for

consumers: self-preferencing and personalised pricing. Both forms of anti-competitive

conduct rely on the operation of powerful algorithms, which in turn rely on the collection

and processing of massive collections of data. Thus, data collection is an issue of

primary importance regarding computationally-driven anti-competitive conduct (Mintz,

2023).

3.1. Self-Preferencing

Dominant platforms utilise algorithms to preference their own products and bury those

of their competitors, a practice known as self-preferencing. This was the abuse of

dominance underlying the European Commission’s (EC) case involving Google

Shopping. The EC found that Google abused its dominance in general search services

in each of the 13 European Economic Area markets, by positioning and displaying its

own products on its general search engine results page more favourably than rivals.5

This raises two detriments:

5
First, the EC found that web pages of rival CSSs could only appear as text-based results in Google’s
SERP, and their SERP ranking was prone to demotion by Google’s algorithm. Second, the EC found that

10
 a. Reduced consumer choice

According to the Commission, Google’s self-preferencing reduced consumer

choice by excluding competing comparison shopping sites (OECD, 2021). By

prioritising its own products in search results and recommendations, Google

controlled the information flow to consumers. While consumers see personalised

suggestions, they do not know the extent to which these recommendations are

influenced by the firm’s self-interest rather than objective relevance. They are

thus exposed to a biassed subset of available products, constraining choice and

leading to skewed purchasing decisions.

In the long term, this would weaken competitors’ sales, forcing them out of the

market and result in greater monopolisation of the incumbent firm. Consequently,

consumers may suffer from reduced choices and higher prices. This underscores

the need for greater regulatory intervention to ensure fair competition.

Thus, Singapore should impose specific rules on online platforms acting as

gatekeepers.6 For instance, regulatory bodies can impose structural remedies

such as functional separation7 to remove self-preferencing incentives. Similarly,

the EU has also suggested the need to consider asymmetric measures for

Google’s own CSS was prominently positioned at the top of the SERP, displayed in richer format and was
not subject to demotion by its own algorithm (OECD, 2021).
6
Gatekeepers are large digital platforms providing any of a predefined set of digital services (‘core
platform services’), such as online search engines, app stores, and messenger services.
7
Functional separation means organisational separation of business units supplying upstream and
downstream services together with associated controls to ensure that the units operate in practice on an
“arm's length” basis.

11
 particular dominant firms, including with respect to the design and operation of

their algorithms (OECD, 2021).

b. Concerns over data privacy

Mass data collection is both a precondition for the creation of algorithms that

enable such forms of anticompetitive conduct, and a negative impact of the

continued operation of these algorithms. The more companies rely on algorithms,

the more they will be incentivised to track consumers to collect data, raising

privacy concerns (Mintz, 2023).

3.2. Personalised Pricing

Personalised pricing occurs when businesses can accurately determine and price what

each customer will pay for a specific product or service (Investopedia, 2023). Today,

personalised pricing is often enabled by algorithms, which are opaque to consumers

and regulators (The Straits Times, 2024). As of 2022, around 40% of firms that have

adopted AI for personalisation use it to set real-time prices and promotions, by

collecting information including consumers’ browsing history, social media activity, or

even their distance from a competitor’s store (Toulouse School of Economics, 2022).

Although personalised pricing can attract more customers by offering lower prices to

those with lower valuations for the product, these consumers benefit little as they have

little willingness to pay in the first place. Simultaneously, those who are willing to pay

more are charged higher prices. This reduces the consumer surplus overall.

12
New forms of AI are on the rise, making it much easier for businesses to carry out

accurate personalised pricing. Reinforcement learning – the ability for AI to learn

through trial and error – is powering personalisation programmes at scale, helping

retailers experiment with new promotion strategies while exploring proven ones via an

automated approach (The Edge, 2024). GenAI, too, is enabling rapid content generation

within personalised campaigns and services. A survey conducted by Boston Consulting

Group showed that 67% of chief marketing officers are exploring GenAI for

personalisation.

However, with new forms of AI, firms’ ability to collect data surreptitiously8 and evade

authority attention has grown significantly. As such, consumers are often unaware of the

vast amounts of information they are giving up to enterprises with potentially malicious

intents. When more consumers share their data, companies may raise prices for those

who choose not to reveal theirs, leading to a new problem where consumers who stay

anonymous face higher prices and suffer unfairly.

Unfortunately, such concerns over the potential misuse of data and data breaches are

currently not explicitly addressed in Singapore’s competition and consumer protection

rules. Hence, there is potential to integrate data protection principles within both the

Competition Act 2003 and the Consumer Protection (Fair Trading) Act (CPFTA) 2004 to

better address issues arising from the misuse of consumer data in competitive

practices, especially in an age of unprecedented AI growth.

8
Today, personalised prices are often concealed as personalised discounts sent by email or smartphone
apps.

13
4. Solutions to Address Data Concerns Regarding Consumers

Given the central role of data collection in the operation of algorithmic self-preferencing

and price personalisation, reforms should be introduced to data privacy rules to mitigate

potential issues of misuse and exploitation. To address these challenges, current

regulations should evolve in two ways:

a. Introducing explicit provisions on data protection in the CPFTA

A new section could be introduced that directly addresses the protection and

rights that consumers have over personal data. Consumers should be licensed to

retract or delete their data at any point in time and the firm should not continue to

use, store or share the data once the request has been made. The section

should also clearly prohibit the misuse of consumer data for unfair practices,

including deceptive personalised pricing.

Additionally, a new dedicated digital platform regulator could be established to

make our current antitrust enforcers more robust (Mintz, 2023). When collecting

information, the firm must make explicit about the type of information collected

and permission should be granted from the consumer.

b. Amending the definition of abuse of dominant position in the Competition Act to

include the control and misuse of consumer data

The definition of a firm’s abuse of dominant position in Section 47 of the

Competition Act should be expanded to include that of data-driven market power,

14
since the possession of vast amounts of consumer data can enable firms to

engage in anti-competitive behaviour. This acknowledges the role of possessing

data in enhancing a firm’s market position, making it explicit that the possession

of private data for strategies that exploit consumer vulnerabilities or promote anti-

competitive behaviour are outlawed.

15
5. Conclusion

The integration of AI into various economic sectors necessitates significant updates to

Singapore’s antitrust and consumer protection laws. While AI has revolutionised various

aspects of commerce, it also introduces unique challenges that current legal

frameworks may be inadequate in addressing. The issues of algorithmic collusion, M&A

facilitated by AI, self-preferencing, and personalised pricing demand a re-evaluation of

existing regulations to ensure that seemingly innocuous yet harmful actions by firms do

not go unchecked.

Drawing from international precedents, Singapore can incorporate explicit provisions for

data privacy in its legal frameworks and redefine certain terms to better encompass

rapidly-evolving market dynamics in this digital epoch. These reforms can reduce legal

ambiguities and better mitigate the risks associated with AI-driven market behaviour.

By adjusting existing concepts to fully encapsulate the scope of AI and establishing

specialised regulatory bodies, Singapore can create a more comprehensive regulatory

environment. Ultimately, robust regulatory measures are essential to prevent the

exploitation of consumer data and maintain a competitive, transparent market.

(2496 words)

16
Bibliography

Bird & Bird. (n.d.). Artificial intelligence: Competition and antitrust's next frontier.

Competition Law Insights. Retrieved from

[Link]

competition-and-antitrusts-next-frontier

Hawkes, C. (2021). The Meteors of the Digital Sky: Algorithms, Cartels, and Other

Antitrust Challenges Posed by Artificial Intelligence. College of Europe Research Paper,

(3). Retrieved from [Link]

paper/ResearchPaper_3_2021_Colm_Hawkes.pdf

Hoff, H. (2022). Algorithmic Collusion as Agreement or Concerted Practice under Article

101 TFEU (Master's thesis). University of Oslo. Retrieved from

[Link]

Algorithmic-Collusion-as-Agreement-or-Concerted-practice-under-Article-101-

[Link]?sequence=1&isAllowed=y

Ezrachi, A., & Stucke, M. E. (2017). The Rise of Behavioural Antitrust. Illinois Law

Review, 2017(5), 1553-1596. Retrieved from [Link]

content/uploads/2017/10/[Link]

17
Marx, L. (2019). Dynamic Pricing in the Age of Artificial Intelligence: Thoughts on

Oligopolistic Coordination. Concurrences, (2). Retrieved from

[Link]

Organization for Economic Cooperation and Development. (n.d.). Exclusionary

Practices and Artificial Intelligence. OECD. Retrieved from

[Link]

Ball, R., & Chappell, L. (2017). Price Discrimination by a Retailer: Should Manufacturers

Worry? Journal of Industrial Economics, 65(1), 97-123. doi:10.1111/joie.12131

Bechtold, S. (2021). The Pitfalls of Pricing Algorithms. Harvard Business Review.

Retrieved from [Link]

Greenberg, S., & Swanson, B. (2022). Data, Algorithms and Antitrust: Next Steps for

Regulators. Lexology. Retrieved from

[Link]

4200270a9ee0

Weizenbaum, J. (2022). Competition Law and Artificial Intelligence - Challenges and

Opportunities. ResearchGate. Retrieved from

[Link]

ntelligence_-_Challenges_and_Opportunities

18
Mayer Brown. (2024). AI Challenges in Competition Law. Mayer Brown. Retrieved from

[Link]

challenges-in-competition-law_mar24.pdf%3Frev=55168f8e10a64e458c3fc1ac7af179d

Smith, J. (2024). The Meteoric Rise of Generative AI. Skadden. Retrieved from

[Link]

meteoric-rise-of-generative-ai

Forum, W. E. (2023). Making Sense of Generative AI: How to Explain Algorithms That

Create. World Economic Forum. Retrieved from

[Link]

Frontier Economics. (n.d.). Assessing Mergers in Innovative Industries. Frontier

Economics. Retrieved from [Link]

insights/articles/article-i10221-assessing-mergers-in-innovative-industries/

Global Competition Review. (n.d.). Self-Preferencing in Digital Markets. Global

Competition Review. Retrieved from [Link]

markets-guide/second-edition/article/self-preferencing-in-digital-markets

19
European Commission. (n.d.). Digital Markets Act: Ensuring fair and open digital

markets. European Commission. Retrieved from [Link]

and-policy/priorities-2019-2024/europe-fit-digital-age/digital-markets-act-ensuring-fair-

and-open-digital-markets_en

Law Insider. (n.d.). Functional separation. Law Insider. Retrieved from

[Link]

separation#:~:text=Functional%20separation%20means%20organisational%20separati

on,an%20%E2%80%9Carm's%20length%E2%80%9D%20basis

Mintz. (2023, December 19). Senators consider reforms to antitrust law to address

digital markets. Mintz. Retrieved from [Link]

center/viewpoints/2191/2023-12-19-senators-consider-reforms-antitrust-law-address

OECD. (n.d.). Policy Options to Address Issues Raised by Digitalisation for Competition,

Consumer Protection and Privacy. OECD. Retrieved from [Link]

[Link]/sites/3acbe1cd-en/[Link]?itemId=/content/component/3acbe1cd-

en#section-d1e9039

Investopedia. (n.d.). What Are the Different Types of Price Discrimination and How Are

They Used? Investopedia. Retrieved from

[Link]

discrimination-and-how-are-they-

20
[Link]#:~:text=First%2DDegree%20Price%20Discrimination,-

In%20a%20perfect&text=Also%20called%20personalized%20pricing%20or,sell%20it%

20for%20that%20price

The Straits Times. (n.d.). Consumers Hate Price Discrimination But They Sure Love a

Discount. The Straits Times. Retrieved from

[Link]

sure-love-a-discount

TSE. (n.d.). Should Prices Be Personalized? TSE. Retrieved from [Link]

[Link]/should-prices-be-

personalized#:~:text=At%20the%20same%20time%2C%20a,even%20though%20it%20

expands%20demand

The Edge Singapore. (n.d.). AI multiplies the value of personalised retail. The Edge

Singapore. Retrieved from [Link]

intelligence/ai-multiplies-value-personalised-retail

21