14. Comparative case law: AI-driven contract systems

The following decisions provide additional authorities on automated transactions, electronic contracting, digital assent, and responsibility for statements made by AI systems. Most predate modern generative AI; their relevance lies in applying established contract principles to transactions conducted through technology.

Case 7: B2C2 Ltd v Quoine Pte Ltd (2019)

Singapore International Commercial Court — [2019] SGHC(I) 3; appeal [2020] SGCA(I) 2

Principle: Automated trading, contractual mistake, and algorithmic transactions.

The dispute arose from cryptocurrency trades executed by algorithms on an electronic trading platform. After unusual price movements, the platform operator reversed the transactions, leading to a dispute over contractual rights and mistake.

The litigation considered how ordinary contractual principles apply when transactions are executed automatically and the relevant parties do not directly intervene in each individual transaction.

Application to AI contracts: A transaction executed by an AI purchasing agent or algorithmic trading system cannot be assessed solely by asking what the software intended. The court must examine the contract, the parties' knowledge and conduct, the relevant legal doctrine, and the circumstances in which the automated transaction occurred.

The case is particularly useful for understanding why automated execution does not necessarily prevent contract formation, while also demonstrating that mistake and other defenses may remain relevant.

Case 8: Chwee Kin Keong v Digilandmall.com Pte Ltd (2005)

Singapore Court of Appeal — [2005] 1 SLR(R) 502

Principle: Online offer and acceptance, electronic transactions, and unilateral mistake.

An online retailer's website displayed printers at an erroneous price. Buyers placed orders, and the dispute concerned whether enforceable contracts had been formed and whether the buyers could take advantage of the pricing error.

The court examined contract formation through an automated website alongside the doctrine of unilateral mistake.

Application to AI contracts: If an AI agent orders goods at a dramatically incorrect price, the legal outcome will depend on the circumstances, including the terms of the transaction, the nature of the error, and the parties' knowledge. Automated acceptance does not necessarily eliminate the possibility of a mistake defense.

The decision is also a reminder that AI systems should include spending limits, price checks, and confirmation procedures for unusually consequential transactions.

Case 9: Thornton v Shoe Lane Parking Ltd (1971)

England and Wales Court of Appeal — [1971] 2 QB 163

Principle: Contract formation through automated systems and notice of contractual terms.

The dispute concerned a motorist who entered an automatic car park and received a ticket from a machine. The court considered when the contract was formed and whether an exclusion clause had been incorporated into it.

The decision illustrates that a contract can arise through an automated transaction and that the timing and adequacy of notice are important to determining which terms bind the parties.

Application to AI contracts: Businesses using AI interfaces must ensure that material contractual terms are communicated clearly at the appropriate stage. A term disclosed only after an AI agent has completed a purchase may not necessarily be incorporated into the contract. The result depends on the applicable law and the facts.

Case 10: Trimex International FZE Ltd v Vedanta Aluminium Ltd (2010)

Supreme Court of India — (2010) 3 SCC 1

Principle: Contract formation through electronic communications.

The dispute concerned a commercial transaction negotiated through electronic communications, including emails. The Supreme Court considered whether the parties had concluded a binding agreement despite the absence of a subsequently executed formal document.

The case supports the proposition that a binding commercial contract can arise through electronic communications when the essential terms and the parties' intention to be bound are established.

Application to AI contracts: An agreement negotiated by an AI system should be examined by reference to the communications, authority granted to the system, essential terms, and applicable rules of contract formation. The absence of a handwritten signature does not by itself establish that no contract exists.

The case did not directly consider AI-generated negotiation or autonomous agents.

Case 11: Register.com, Inc. v Verio, Inc. (2004)

United States Court of Appeals, Second Circuit — 356 F.3d 393

Principle: Contractual assent through repeated conduct and notice of terms.

The dispute concerned a company that repeatedly accessed information through a website after receiving notice of restrictions governing the use of that information.

The court considered the significance of notice and repeated conduct in determining whether contractual restrictions were binding.

Application to AI contracts: When an AI agent repeatedly uses a service through an API or website, a court may need to determine whether the principal's conduct, the terms presented, and the circumstances establish assent. The mere technical ability to access a website should not be treated as conclusive proof that every term was accepted.

This is especially relevant where an AI agent accepts terms automatically without a human reviewing each interaction.

Case 12: Moffatt v Air Canada (2024)

British Columbia Civil Resolution Tribunal — 2024 BCCRT 149

Principle: Responsibility for misleading information supplied by an airline's chatbot.

A passenger relied on information provided by Air Canada's chatbot about bereavement fares. The airline subsequently disputed the passenger's entitlement to the fare treatment described by the chatbot.

The tribunal found Air Canada liable for negligent misrepresentation in the circumstances. The case demonstrates that a business may be held responsible for misleading information provided through its customer-facing automated system.

Application to AI contracts: A company cannot necessarily avoid responsibility for inaccurate representations simply by attributing them to a chatbot. Depending on the applicable law, a business may face liability where a customer reasonably relies on information supplied through its official AI interface.

The case concerned a chatbot and misrepresentation, not the general validity of contracts autonomously negotiated by AI.

These additional decisions reinforce a consistent comparative principle: existing contract law can often address technology-mediated transactions, but the outcome depends on authority, assent, notice, mistake, and the legal consequences of the system's conduct.

15. Indian statutory framework for AI-driven contract systems

India does not have a single comprehensive statute specifically governing all contracts negotiated or concluded by autonomous AI agents. Existing legislation provides a foundation, but questions concerning advanced AI autonomy, authority, unexpected outputs, and allocation of liability remain fact-dependent.

The principal legal provisions include:

Indian Contract Act, 1872: Section 10 addresses the requirements of valid contracts; Sections 13–19 address consent, free consent, and certain grounds for avoiding agreements; Sections 182 onwards govern agency.

Information Technology Act, 2000: Section 10A recognizes that a contract is not unenforceable solely because electronic means were used in its formation.

Information Technology Act, 2000, Section 11: Provides rules for attributing electronic records to their originator, including certain automatically operating information systems.

Consumer Protection Act, 2019: May apply to misleading representations, unfair trade practices, and other consumer disputes involving AI-powered services.

Digital personal data protection legislation: Applicable requirements must be considered where an AI contract system processes personal data.

The central question is whether the actions of an AI system can be attributed to the individual or business that deployed it. An AI system does not automatically acquire independent legal personality simply because it can negotiate, make recommendations, or execute transactions.

For example, if a company authorizes an AI procurement system to purchase goods within a specified budget, a transaction within that authority may bind the company under applicable law. If the system unexpectedly agrees to a substantial additional obligation, the result will depend on the scope of authority, the contract, the conduct of the parties, and relevant agency and electronic-transactions principles.

There is not yet a comprehensive body of Indian appellate case law specifically determining the validity of contracts concluded independently by modern generative AI agents. The established electronic-contract decisions discussed above provide useful analogies, but their principles must be applied carefully.

16. Practical model for responsible AI contracting

A legally sound AI contract system should incorporate the following controls:

Authority controls

Define what the AI may negotiate, purchase, amend, accept, or reject, together with spending limits and restricted transactions.

Contract validation

Verify the identity of the contracting party, essential terms, applicable law, acceptance, and any required formalities before execution.

Audit trails

Maintain reliable records of instructions, offers, counteroffers, approvals, acceptance, and material system actions.

Human approval for high-risk commitments

Require authorized human approval for major acquisitions, guarantees, borrowing, substantial liability, and unusual contractual deviations.

17. Conclusion

Comparative civil law indicates that AI-driven contract systems should generally be analyzed through established principles of contract formation, consent, authority, mistake, misrepresentation, agency, and civil liability.

The twelve decisions discussed in this explanation offer a foundation for that analysis. In particular, B2C2 v Quoine, Chwee Kin Keong v Digilandmall, Thornton v Shoe Lane Parking, Trimex v Vedanta, Register.com v Verio, and Moffatt v Air Canada illustrate how legal principles can apply to automated transactions, electronic communications, digital terms, and AI-mediated representations.

These decisions do not establish that every AI-generated contract is valid or that every unexpected AI action binds its operator. Instead, they demonstrate why courts must examine the agreement, the parties' conduct, the authority granted, and the relevant statutory framework.

The most defensible approach is to recognize that AI can facilitate the formation and performance of contracts without treating the technology as an independent legal person. Businesses deploying AI should establish clear authority limits, preserve reliable transaction records, verify material representations, protect personal and confidential information, and provide human oversight for high-consequence commitments.

Ultimately, the central objective of comparative AI contract law is to preserve contractual certainty and commercial innovation while ensuring that responsibility for AI-mediated transactions remains legally identifiable and enforceable.