Enforceability of AI-Drafted Contracts in India: A Legal Analysis

This Blog is Written by Devanshu Singh, 4th Year, BA.LLB, Lloyd Law College, Greater Noida.

Blog 17 | Edition VII

Introduction

Generative artificial intelligence (AI) tools are now being leveraged by law firms, start-ups, and even individual founders to generate Non-Disclosure Agreements, vendor agreements, and employment contracts, with minimal human intervention. Surveys indicate that legal-tech adoption in India, especially by in-house counsel, for AI-driven drafting is rising, as small firms look to reduce costs. But can such a contract be enforced in a court of law? Not a single Indian court (or indeed, any other court in the world) has ruled on the enforceability of a contract either drafted by an AI or substantially negotiated by an AI. This renders the policy issue around the enforcement of AI-drafted contracts a matter of speculation in legal-AI circles. This blog argues that AI-drafted contracts can indeed be enforced by Indian courts, provided certain conditions are met, such as evidence of human intent to be bound by the contract, authenticity of the electronic record, and free consent to its terms.

Validity of an Electronic Contract

The Indian Contract Act, 1872 does not draw a distinction between a contract that is handwritten or not. The Information Technology Act, 2000 and its Section 10A relating to electronic commerce does not make any exception either for contracts formed through agents, or AI systems. Thus, there is no ground to argue that an agreement is not enforceable simply on the grounds of it being electronic. However, admissibility of electronic records as evidence is governed by section 65B of the Indian Evidence Act, 1872 (replaced by section 63 of Bharatiya Sakshya Adhiniyam, 2023), which requires a certificate of authenticity for an electronic record. A digital signature as per the IT Act, 2000 permits a record to be attributed to a signatory. Thus, there is no hurdle, from a legal standpoint, to an electronic contract being enforced. This blog now turns to examining the validity and enforceability in law of a contract whose contents have been created by an AI system.

Human Agency and Intent

A contract traditionally has two human parties coming together in a “meeting of minds”. When the machine is generating contract clauses, it is vital to question whether a human party to a contract that has given assent to an indemnity clause generated by an AI tool can be said to have consented to the proposal as “proposal” as defined under Section 4 of the Indian Contract Act, 1872? When the onus is on an individual to read and accept a machine’s terms, can such an acceptance be considered to meet the test of free consent as per Section 14 of the Act, or does it fall foul of a unilateral mistake by the human being in accepting a contract?

Authenticity of Electronic Records: Certificate of Authenticity Versus AI Authorship

When we accept a digital certificate as authentication of records per Section 65B of the Indian Evidence Act, 1872, we are essentially accepting that the communication is authorized by a human on the other end. But when the contracts are being created with the assistance of AI, authenticity of records could be questioned beyond the mere validity of the electronic signature. Courts will have to examine whether an AI tool is a “word processor” or has crossed the Rubicon into “agent”. If there is human consent to be bound by the terms drafted by the machine, does it violate the requirements for a certificate of authenticity under Section 65B of the Indian Evidence Act, 1872, or is the use of artificial intelligence as an agent a novel development in contracts?

Comparison with International Law and Potential Remedies

The Uniform Electronic Transactions Act (UETA) and the E-Sign Act of the US have already addressed the nuances mentioned above and offer potential remedies which may be considered by India. When an electronic agent is used to make an offer or acceptance, the contract is deemed to be formed by the human agent who deployed it, and the law does not require that a human had to read and understand every part of the document for it to be valid. The EU is taking a different approach by regulating AI systems and their impact on safety and fundamental rights, while the Model Law on Electronic Commerce of the UNCITRAL also provides that automated messaging systems may form valid contracts so long as the obligations under it are attributable to the human sender. The Information Technology Act, 2000 does not, however, contain any equivalent of UETA’s Section 403 regarding “electronic agents”, leaving Indian courts to grapple with the enforceability of AI-drafted contracts on a discretionary basis.

Possible Solutions

Amending the Information Technology Act: India may wish to consider amending the IT Act along the lines of UETA’s Section 403, regarding obligations of an “electronic agent”. An AI tool used by humans to draft contracts should be deemed to have “authorised” the information contained within, so long as the human deploying it had a “reasonable opportunity” to review the information within it. Adopting “human-in-the-loop” mechanisms for high-risk clauses with a check for free consent: A “human-in-the-loop” review mechanism should be mandatory for high-risk clauses such as indemnifications or arbitration clauses. Instead of a human "accepting" a contract "as is", they should be able to "opt-in" to specific things in a contract and agree to those specifically. This could be a method of checking free consent, particularly if a party has accepted standard, or boilerplate, terms generated by the AI. A more nuanced approach should be taken when it comes to contract formation using AI, wherein the log of a record’s drafting process may be used to establish the “authenticity” of a record for purposes of the Evidence Act, instead of the certificate of authenticity that is standard practice at present.

Conclusion

AI-drafted contracts are not inherently unenforceable by virtue of being drafted by a machine. Courts will enforce such contracts if human consent to be bound by the terms is genuine, and the record containing the information is authentic. The blog ends with suggestions on how India can render AI-drafted contracts enforceable by amending the Information Technology Act, putting in place “human-in-the-loop” reviews, and revising the Evidence Act regarding the admissibility of AI records as evidence.

(Write to the author at devanshusingh2906@gmail.com.)