Revision notes

AI Ethics, Safety and Law: Foundation

The key points, rules to remember, tables and diagrams from all 5 lessons, on one page for quick revision. To keep a copy, use your browser's Print and choose "Save as PDF".

Lesson 1

Why AI ethics matters: the core principles of responsible AI

Why AI needs principles of its own

  • Scale One flawed rule can affect lakhs of people in a day.
  • Learns from data It can quietly copy unfair patterns from the past.
  • Hard to explain Even its builders cannot always say why one answer came out.
Five principles of responsible AI
PrinciplePlain meaningQuestion to ask
FairnessThe system should not treat people worse because of caste, religion, gender, region or similar traitsDoes it give worse results to some groups?
AccountabilityA named person or organisation answers for what the system doesWho is responsible if it goes wrong?
TransparencyPeople are told that AI is being used, broadly how, and with what dataDo users know an AI is involved?
ExplainabilityThe main reasons behind a particular decision can be given in understandable termsCan the affected person be told why?
Human oversightA person can check, correct or stop the system, especially for serious decisionsCan a human step in?

Responsible AI through the whole life of a system

  1. Decide whether to use AI
  2. Collect data
  3. Build and test
  4. Launch
  5. Keep watching in use

The more an AI decision can harm a person, the more fairness, explanation and human review it needs. The machine assists; a person stays responsible.

Key takeaways: AI raises new ethical questions because it works at scale, learns from data and can be hard to explain. The five core principles are fairness, accountability, transparency, explainability and human oversight. Responsible AI means applying them throughout a system's life, with safeguards that grow with the seriousness of the decision.

Read the full lesson

Lesson 2

Bias and fairness: how unfair outcomes enter AI systems and how to reduce them

Bias in, bias out

  1. Unfair data or design choices Data from the real world and choices made by people.
  2. The AI learns the pattern
  3. Unfairness repeated at scale The system can repeat it at very large scale.
Common sources of bias in AI
Source of biasWhat happensSimple illustration
Historical biasThe data faithfully records an unfair pastPast hiring favoured one group, so the model learns to favour it
Representation (sampling) biasSome groups are missing or too few in the dataA face dataset with few photos of darker-skinned people
Measurement or label biasWhat is recorded is a poor stand-in for what mattersUsing 'number of complaints' as a measure of an employee's quality
Proxy variablesA harmless-looking feature closely tracks a protected traitResidential pin code standing in for community or income
Feedback loopsThe model's own outputs shape the next round of dataMore checks in one area find more cases there, which leads to even more checks
Design choicesThe goal the builders set ignores fairnessOptimising only for overall accuracy, ignoring errors for small groups

An average can hide an unfair gap

  • Group P
    • 10 wrong matches in 1,000 photos
    • Error rate 1%
  • Group Q
    • 50 wrong matches in 1,000 photos
    • Error rate 5% - five times that of Group P
  • Both groups together
    • 60 wrong matches in 2,000 photos
    • Error rate 3%, which sounds acceptable

Bias in, bias out. A system that is accurate on average can still be unfair to a group - always measure results separately for each group.

Key takeaways: bias enters through unfair historical data, missing groups, poor labels, proxy variables, feedback loops and design goals. It shows up in hiring, lending and face recognition. Detect it by comparing rates across groups; reduce it with better data, proxy checks, diverse teams, human review and regular audits.

Read the full lesson

Lesson 3

Privacy and data: consent, the DPDP Act, 2023 and the right to privacy

IdeaMeaningEveryday example
Personal dataAny data about a person who can be identified from itName, mobile number, Aadhaar number, photo, location
ConsentThe person's free and informed agreement to a specific use of their dataTicking a box you have read, for a clearly stated purpose
Purpose limitationData collected for one purpose is used only for that purposeA phone number taken for login is not sold for marketing
Data minimisationCollect only the data actually neededA quiz app does not need your contacts or gallery
Storage limitationKeep data only as long as neededDelete records once the purpose is over
SecurityProtect data against leaks and theftEncryption, access controls, strong passwords
TermWho it is
Data principalThe individual to whom the personal data relates (for a child, this includes the parent or lawful guardian)
Data fiduciaryThe person or organisation that decides the purpose and means of processing personal data
Data processorOne who processes personal data on behalf of a data fiduciary
Consent managerA registered body through which a data principal can give, manage, review and withdraw consent
Data Protection Board of IndiaThe body set up under the Act to handle complaints and breaches

Two laws, two jobs

  • IT Act, 2000
    • Deals broadly with the digital world and cyber crime
    • Legal recognition for electronic records and digital signatures
    • Offences such as hacking, identity theft and cheating by impersonation
    • Governs intermediaries such as social media platforms
  • DPDP Act, 2023
    • Deals specifically with digital personal data
    • Consent must be free, specific, informed, unconditional and unambiguous
    • Rights to information, correction, erasure and grievance redressal
    • Data Protection Board of India handles complaints and breaches

Who is who: the coaching institute example

  1. Data principal Each student - the person the data is about.
  2. Data fiduciary The institute, which decides why and how the data is used.
  3. Data processor The cloud company, which processes the data on the institute's behalf.

Data principal = the person; data fiduciary = decides why and how; data processor = acts on the fiduciary's behalf. Collect only what is needed, only for the purpose agreed.

Key takeaways: privacy is a fundamental right (Puttaswamy, 2017, Article 21). The DPDP Act, 2023 governs digital personal data through consent, purpose limits and duties on data fiduciaries, with a Data Protection Board. The IT Act, 2000 covers electronic records and cyber offences. For AI, this means using personal data only with valid consent, only as much as needed, and keeping it secure.

Read the full lesson

Lesson 4

Deepfakes, misinformation and cyber safety: spotting and reporting AI-enabled fraud

TermMeaning
DeepfakeAudio, video or an image created or altered with AI so that a person appears to say or do something they never did
Voice cloneAn AI copy of a person's voice, made from samples of their speech
MisinformationFalse or misleading information shared without intent to deceive
DisinformationFalse information created or spread deliberately to deceive
Synthetic mediaAny media generated by AI, whether used honestly or not

Three warning signs of AI-enabled fraud

  • Urgency Act now.
  • Secrecy Do not tell anyone.
  • A demand For money, an OTP or personal details.
These signs matter more than how real the voice or face looks.

If you have lost money to cyber fraud

  1. Call 1930 quickly Early reporting improves the chance the money can be traced and blocked.
  2. Report on cybercrime.gov.in The National Cyber Crime Reporting Portal.
  3. Inform your bank
  4. Keep the evidence Screenshots, phone numbers, transaction IDs and messages; do not delete the chat.

Urgency + secrecy + money = stop and verify through another channel. Lost money? Call 1930 and report on cybercrime.gov.in without delay.

Key takeaways: deepfakes and voice clones make it possible to fake faces and voices convincingly. Judge a message by its demands, not by how real it looks. Verify through a second channel, never share OTPs, and report financial cyber fraud quickly on 1930 or cybercrime.gov.in.

Read the full lesson

Lesson 5

AI safety, jobs and governance: human-in-the-loop and why people stay responsible

Two groups of AI safety problems

  • Accidents Harms nobody intended.
    • Confident but wrong answers
    • Fails on situations it was not trained for
    • Tricked by unusual inputs
  • Misuse Deliberate harm by people.
    • Cloning voices for fraud
    • Writing convincing scam messages
    • Spreading disinformation
    • Attacking computer systems
TypeExampleTypical safeguard
Accident: errorA chatbot states a wrong fact confidentlyCheck important facts against reliable sources
Accident: new situationA crop-advice tool trained on one region fails in anotherTest in each setting before use
Misuse: fraudVoice cloning to impersonate a bank officerVerification steps, awareness, law enforcement
Misuse: disinformationMass-produced fake news postsLabelling, platform action, fact-checking
LevelHow it worksSuitable for
Human-in-the-loopThe AI suggests; a person reviews and approves before anything happensSerious decisions: benefits, loans, medical treatment, legal matters
Human-on-the-loopThe AI acts on its own while a person monitors and can step inFast, lower-risk tasks such as filtering spam
Human-out-of-the-loopThe AI acts with no human involvementOnly very low-risk tasks

Match the human control to the risk

  • Sorting by completeness
    • Low risk
    • AI flags missing documents
    • A person monitors: human-on-the-loop
  • Rejecting an application
    • High risk
    • An officer reads the case and decides: human-in-the-loop
    • The student gets reasons and a way to appeal

AI can recommend; a person decides and answers for the decision. The higher the risk, the closer the human must stay to the loop.

Key takeaways: AI safety covers both accidents and misuse. Automation changes routine tasks first, so reskilling matters. Serious decisions need a human in the loop who genuinely reviews. Governments use risk-based laws, existing laws with guidance, and international principles. In every approach, people remain accountable.

Read the full lesson

Privacy · Terms · Refunds · Delivery