Law in the Age of AI
Contracts, copyright, data and accountability: a practical map of the questions artificial intelligence is forcing lawyers — and everyone who signs contracts — to answer.
Law has met new technology before — the printing press, the telephone, the internet, the smartphone. Each time the same pattern played out: the technology raced ahead, the rules struggled to keep up, and in the gap between the two, lawyers were asked to explain what everyone was allowed to do. Artificial intelligence and machine learning are the latest chapter, and arguably the fastest.
I spend most of my working life on the transactional side of law — due diligence, M&A, contracts — and my curiosity about Web 3.0 and AI has grown out of that daily work. Because when a new technology matters, it does not arrive in a white paper. It arrives in a clause, a vendor agreement, a data room. So here is a map of where AI is already pressing on the law, organised by the documents and decisions that lawyers actually see.
1. Contracts: who carries the risk?
Most AI risk today is being allocated, quietly, in contracts. If you are buying or building AI-enabled tools, a handful of questions decide who bears the downside:
- Outputs and liability. If the tool produces a wrong, biased or infringing result, who is responsible — the vendor, the deployer or the user? Watch for broad disclaimers paired with narrow indemnities.
- Training and data rights. May the vendor use your inputs to improve its models? Is that opt-in or opt-out, and can it be reversed?
- Confidentiality. Where does sensitive data travel, who can see it, and how long is it retained?
- Human oversight. Is a person accountable for high-stakes decisions, or is the tool quietly making them?
The most consequential AI regulation this year may be the indemnity clause on page fourteen.
2. Intellectual property: creators, authors and training data
Intellectual property is where my interest in the field began, and it is where the hardest philosophical questions sit. Who is the author of a work that a person shaped through prompts but a model generated? Indian copyright law, notably, defines the author of a computer-generated work as the person who causes the work to be created — a definition written long before generative AI, and one that sits awkwardly with today’s tools.
The second question is even more contested: can copyrighted material be used to train models without permission or payment? Courts in India and abroad are being asked precisely this, and the answers will shape both the creative economy and the AI industry. Until they arrive, the sensible approach for businesses is practical: know where your training data came from, keep records, and negotiate warranties and indemnities accordingly.
3. Data protection and privacy
AI systems are hungry for data, and data protection law is the natural counterweight. India’s Digital Personal Data Protection Act, 2023 puts consent, purpose limitation and accountability at the centre of how personal data may be processed. For any company deploying AI, this raises unglamorous but crucial questions: Was consent properly obtained? Is the data being used for the purpose it was collected for? Can a person ask what is being done with their information?
4. Regulation: risk-based, or not yet?
Globally, approaches differ. The European Union has adopted a risk-based AI Act that places heavier obligations on higher-risk uses. India has so far leaned largely on existing laws — the Information Technology Act, data protection rules and sector regulators — rather than a single AI statute. For cross-border businesses this patchwork is a compliance puzzle; for policymakers it is a live design question about how to encourage innovation without leaving individuals unprotected.
5. Web 3.0: the same questions, a different ledger
Blockchain-based systems raise a sibling set of issues: how to enforce an agreement that executes automatically, who is accountable in a decentralised structure, and how tokenised assets fit within securities, tax and consumer law. The technology changes; the lawyer’s questions — who owes what to whom, and how do we prove it — stay remarkably constant.
A five-point checklist before you sign an AI agreement
- Map exactly what data goes in, where it is stored and who can access it.
- Get clear, written answers on whether your inputs are used to train models.
- Negotiate liability and indemnities for inaccurate, biased or infringing outputs.
- Keep a human accountable for consequential decisions, and document the review.
- Diarise a review: the law, the tool and your use case will all change within a year.
The lawyer’s new job
There is a fear that AI will replace lawyers. I think a more useful question is what kind of lawyer thrives beside it. Routine drafting and first-pass review will increasingly be automated; judgement, negotiation, ethics and the ability to translate between engineers, executives and regulators will matter more, not less.
The task in front of us is to be translators — to help businesses adopt these tools with their eyes open, to help creators keep the value of their work, and to help regulators write rules that protect people without freezing progress. It is a moment that calls for curiosity over panic, and I count myself lucky to be practising in it.
This article is general commentary for information only and is not legal advice. Laws and court decisions in this field are developing quickly; please consult a qualified lawyer for advice on your specific situation.