Module-2

AI Risk & Ethics

Topics to be covered
• Privacy
• Bias
• Responsible AI principles
AI Risk & Ethics: Privacy
• Privacy in Artificial Intelligence (AI) refers to the protection of an
individual's personal, sensitive, and confidential information
throughout the entire AI lifecycle.
• AI systems often rely on large volumes of data to learn, make
predictions, and support decision-making.
• Privacy ensures that this data is collected, stored, processed, and
shared securely, ethically, and only for legitimate purposes.
Why is Privacy Important in AI?
• Privacy is one of the most critical ethical concerns in AI because AI systems
frequently process highly sensitive information, such as medical records,
financial transactions, biometric data, and online activities.

Importance of Privacy
•Protects users' personal and confidential information.
•Prevents identity theft, fraud, and cyberattacks.
•Builds trust between users and AI-powered applications.
•Ensures compliance with national and international privacy laws.
•Protects sensitive information such as healthcare, banking, and educational records.
•Encourages responsible and ethical AI development.
•Reduces the risk of data misuse and unauthorized surveillance.
Privacy Risks in AI

1. Excessive Data Collection

Many AI applications collect more data than necessary to improve prediction accuracy.
Collecting unnecessary personal information increases the risk of privacy violations.
Example:
A shopping application requests access to a user's contacts, location, microphone, and
photos even though only the location is required for delivery.
Impact:
•Increased privacy risk
•Greater chance of data leakage
•User distrust
Privacy Risks in AI (Cont…)
2. Data Collection Without Consent
Some AI systems collect or analyze personal data without obtaining clear and
informed user consent.
Example:
A mobile application continuously tracks a user's location in the background without
clearly informing them.
Impact:
•Violation of user privacy
•Legal and ethical concerns
•Loss of customer confidence
Privacy Risks in AI (Cont…)
3. Data Breaches
AI systems often store large amounts of sensitive information. Weak security
measures or cyberattacks can expose this data to unauthorized users.
Example:
A hospital's AI system storing patient records is hacked, resulting in the exposure of
confidential medical information.
Impact:
•Financial losses
•Identity theft
•Damage to organizational reputation
Privacy Risks in AI (Cont…)
4. Unauthorized Data Sharing
Organizations may share personal information with advertisers, business partners,
or third-party organizations without explicit user permission.
Example:
A social media platform shares user browsing history with advertising companies to
deliver personalized advertisements.
Impact:
•Loss of user trust
•Ethical violations
•Regulatory penalties
Privacy Risks in AI (Cont…)
5. Identity Theft
Cybercriminals can misuse leaked personal information to impersonate individuals
and commit fraud.
Example:
A hacker obtains a person's Aadhaar number, bank details, and phone number to
perform unauthorized financial transactions.
Impact:
•Financial fraud
•Account takeover
•Legal complications
Privacy Risks in AI (Cont…)
6. Mass Surveillance
AI-powered surveillance systems can continuously monitor people's activities using
cameras, facial recognition, and location tracking.
Example:
Smart city surveillance cameras use facial recognition technology to identify
individuals in public places.
Benefits
•Improved public safety
•Crime prevention
Risks
•Loss of anonymity
•Potential misuse of surveillance data
•Violation of civil liberties
Privacy Risks in AI (Cont…)
7. Re-identification of Anonymous Data
Even after removing personal identifiers, advanced AI techniques can sometimes
combine multiple datasets to identify individuals.
Example:
Anonymous health records are linked with publicly available demographic
information to identify specific patients.
Impact
•Loss of anonymity
•Privacy violations
Real-World Examples
Healthcare
AI helps doctors diagnose diseases, recommend treatments, and monitor patients.
Privacy Risk
Medical records contain highly sensitive information. Unauthorized access can
expose patients' health conditions.
Example
AI systems analyzing MRI or CT scans must securely store patient records and
ensure that only authorized medical professionals can access them.
Real-World Examples (Cont…)
Banking and Finance
Banks use AI for fraud detection, loan approvals, and personalized financial
services.
Privacy Risk
Financial information such as account numbers, transaction history, and credit
scores must remain confidential.
Example
AI detects unusual credit card transactions while ensuring customer financial data
is encrypted and securely processed.
Real-World Examples (Cont…)
E-commerce
Online shopping platforms use AI to recommend products and improve customer
experience.
Privacy Risk
Purchase history, browsing behavior, and payment information may be collected for
recommendation systems.
Example
Amazon recommends products based on previous purchases while protecting
customer payment information using encryption.
Real-World Examples (Cont…)
Social Media
AI personalizes news feeds, advertisements, and content recommendations.
Privacy Risk
AI collects user interests, messages, photos, and online activities, which may create
detailed personal profiles.
Example
Facebook and Instagram recommend content based on user interactions while
following privacy policies and user consent settings.
 Methods to Protect Privacy in AI
Organizations should adopt strong privacy-preserving techniques throughout the AI lifecycle.

Technical Measures Privacy-Preserving AI Techniques

•Data Encryption •Data Anonymization
•Secure Cloud Storage •Pseudonymization
•Multi-factor Authentication (MFA) •Generalization
•Access Control Mechanisms
•Secure APIs (ID: P001 | Age: 25 | Phone: XXXX |
•Continuous Security Monitoring City: Bhubaneswar)

Organizational Measures
•Obtain informed user consent.
•Collect only necessary information (Data Minimization).
•Maintain transparency regarding data usage.
•Conduct regular privacy audits.
•Train employees on data protection practices.
•Follow privacy-by-design principles.
Advantages of Protecting Privacy
• Builds user trust and confidence.
• Encourages responsible AI adoption.
• Protects sensitive personal information.
• Reduces the likelihood of cyberattacks and identity theft.
• Ensures compliance with legal and ethical standards.
• Enhances the reputation of organizations using AI.
• Promotes fairness and transparency in AI systems.

Privacy Laws and Regulations

Regulation Region Purpose
GDPR(General Data Protection Regulation) European Union Protects personal data and privacy rights
DPDP Act, 2023 (Digital Personal Data Protection) India Governs the processing of digital personal data
CCPA(California Consumer Privacy Act) California, USA Gives consumers greater control over personal information
AI Risk & Ethics: Bias
• AI Bias refers to the unfair or prejudiced behavior of an AI system
that results in unequal treatment or inaccurate decisions for certain
individuals or groups.
• Bias usually occurs when the data used to train the AI model is
incomplete, unbalanced, or reflects existing human prejudices.
AI Risk & Ethics: Bias (cont..)
Why is Bias an Ethical Concern?
Bias is one of the biggest ethical challenges in AI because AI systems are
increasingly used in hiring, healthcare, banking, education, law
enforcement, and many other critical areas. Biased AI can lead to
discrimination and unfair decision-making.

Importance of Addressing Bias

•Ensures fairness in AI decisions.
•Prevents discrimination against individuals or groups.
•Increases trust in AI systems.
•Improves decision accuracy.
•Supports responsible and ethical AI development.
•Ensures compliance with legal and ethical standards.
AI Risk & Ethics: Bias (cont..)
Types of AI Bias
1. Data Bias
It occurs when the training data is not representative of the real-world population.
Example:
A facial recognition system trained mostly on images of adults may perform poorly when identifying children or elderly
people.
2. Algorithmic Bias
It occurs when the AI algorithm unintentionally favors one group over another due to its design or optimization.
Example:
A loan approval model gives lower approval rates to applicants from certain neighborhoods because of patterns in
historical data.
3. Human Bias
Occurs when human beliefs or prejudices influence AI system development.
Example: Developers unintentionally label training data based on personal opinions or stereotypes.
4. Sampling Bias
Occurs when some groups are underrepresented or overrepresented in the dataset.
Example: A medical AI system trained only on data from urban hospitals may not perform well for rural populations.
AI Risk & Ethics: Bias (cont..)
Real-World Examples
1. Hiring and Recruitment
AI systems screen resumes to shortlist candidates.
Bias Risk:
If historical hiring data favors one gender or educational background, the AI may unfairly reject qualified candidates.

2. Healthcare
AI assists doctors in disease diagnosis.
Bias Risk:
A model trained mainly on one ethnic group may perform less accurately for other populations.

3. Banking and Finance

AI is used for credit scoring and loan approval.
Bias Risk:
Historical lending practices may cause AI to unfairly deny loans to certain groups.

4. Education
AI evaluates student performance and recommends learning resources.
Bias Risk:
Students from different educational backgrounds may receive unfair assessments if the training data is unbalanced.
AI Risk & Ethics: Bias (cont..)
Causes of AI Bias

•Biased or incomplete training data.

•Lack of diversity in datasets.
•Human prejudice during data collection or labeling.
•Poor feature selection.
•Algorithm design choices.
•Lack of regular testing and monitoring.
•Insufficient representation of minority groups.
.
AI Risk & Ethics: Bias (cont..)
How to Reduce AI Bias?

Organizations can reduce AI bias by adopting fair and transparent AI

development practices.
•Regularly test AI models for fairness.
•Remove sensitive attributes when appropriate (e.g., race, religion, gender)
if they are not relevant to the task.
•Increase diversity in AI development teams.
•Continuously monitor AI systems after deployment.
AI Risk & Ethics: Responsible AI Principles
• Responsible AI refers to the design, development, deployment, and use of Artificial
Intelligence systems in a way that is ethical, fair, transparent, secure, accountable, and
beneficial to society. It ensures that AI technologies respect human rights, protect privacy,
avoid discrimination, and are used safely.

• Responsible AI means developing and using AI systems that are fair, transparent, secure,
accountable, and aligned with human values.
AI Risk & Ethics: Responsible AI Principles (cont…)
Why is Responsible AI Important?
• As AI becomes more powerful and widely used, it can significantly impact people's lives.
Responsible AI ensures that these technologies are used safely and ethically.
Importance of Responsible AI
• Ensures fairness in AI decisions.
• Protects user privacy and data.
• Builds trust in AI systems.
• Reduces bias and discrimination.
• Improves transparency and accountability.
• Supports legal and ethical compliance.
• Promotes sustainable and human-centered AI.
Core Principles of Responsible AI
1. Fairness
AI systems should treat all individuals and groups equally without
discrimination based on gender, race, age, religion, or other protected
characteristics.
Example:
A recruitment AI should evaluate candidates based on skills and
qualifications rather than gender or ethnicity.
Benefits
•Equal opportunities
•Reduced discrimination
•Better public trust
Core Principles of Responsible AI (cont…)
2. Transparency
Users should understand how AI systems make decisions and how their
data is used.
Example:
A bank should explain why an AI system approved or rejected a loan
application.
Benefits
•It builds user confidence
•It is easier to identify errors
•It improves accountability
Core Principles of Responsible AI (cont…)
3. Accountability
Organizations and developers must take responsibility for the decisions
made by AI systems.
Example:
If an autonomous vehicle causes an accident, the manufacturer and
developers should investigate and address the issue.
Benefits
•It Encourages ethical AI development
•It Improves reliability
•It Supports legal compliance
Core Principles of Responsible AI (cont…)
4. Privacy and Security
AI systems should protect users' personal information and prevent
unauthorized access.
Example:
Healthcare AI systems should encrypt patient records and allow access
only to authorized medical professionals.
Benefits
•It Protects sensitive information
•It Prevents cyberattacks
•It Increases user trust
Core Principles of Responsible AI (cont…)
5. Reliability and Safety
AI systems should operate consistently, accurately, and safely under
different conditions.
Example:
A medical AI system should be thoroughly tested before assisting doctors
in diagnosing diseases.
Benefits
•Reduces errors
•Improves system performance
•Protects human lives
Responsible AI (Real-World Examples)
Healthcare Social Media
Responsible AI Practices Responsible AI Practices
•Secure patient records •Detect harmful content
•Explain AI-based diagnoses •Protect user privacy
•Human review of treatment recommendations •Reduce misinformation and biased recommendations

Banking
Responsible AI Practices
•Fair loan approval Autonomous Vehicles
•Explain credit decisions Responsible AI Practices
•Protect customer financial data •Safe driving decisions
•Continuous monitoring
•Human intervention during emergencies
Education
Responsible AI Practices
•Fair student assessments
•Transparent grading systems
•Protection of student information
Reference
• [1] S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 4th
ed. Hoboken, NJ, USA: Pearson, 2021.
• [2] UNESCO, Recommendation on the Ethics of Artificial Intelligence. Paris,
France: UNESCO, 2021.
• [3] National Institute of Standards and Technology (NIST), Artificial
Intelligence Risk Management Framework (AI RMF 1.0). Gaithersburg, MD,
USA: NIST, Jan. 2023.
• [4] European Commission, Ethics Guidelines for Trustworthy AI. Brussels,
Belgium: High-Level Expert Group on Artificial Intelligence, 2019.
• [5] B. Mittelstadt, P. Allo, M. Taddeo, S. Wachter, and L. Floridi, "The ethics
of algorithms: Mapping the debate," Big Data & Society, vol. 3, no. 2, pp.
1–21, 2016.