Patent Enablement for Machine-Generated Compounds

Patent enablement becomes particularly important when artificial intelligence (AI) systems generate new chemical compounds, pharmaceutical candidates or molecular structures. The central question is whether a patent application discloses enough information for a person skilled in the relevant field to make and use the claimed compounds without undue experimentation. The fact that a compound was generated by a machine does not, by itself, establish patentability or satisfy the disclosure requirement.

Legal Framework in India

Under Section 10(4) of the Patents Act, 1970, a complete specification must fully and particularly describe the invention, explain how it operates or is used, disclose the method of performing it, and provide the best method known to the applicant. Section 10(5) further requires clear, succinct claims fairly based on the disclosure.

Government of India

For machine-generated compounds, adequate disclosure may require chemical structures, synthesis procedures, relevant experimental data, reproducible testing methods and evidence supporting the claimed properties. If an AI model predicts thousands of compounds but the application provides only a few examples without a reliable method of obtaining the rest, the broader claims may fail the enablement requirement.

The invention must also satisfy novelty and inventive-step requirements under Section 2(1)(j) and Section 2(1)(ja). Section 3(d) may restrict certain claims involving new forms of known substances unless the statutory requirements are met. AI's role in generating a compound does not remove these requirements.

Important Case Laws

1. Amgen Inc. v. Sanofi, 598 U.S. 594 (2023)

The US Supreme Court held that Amgen's broad claims to a class of antibodies binding to PCSK9 were insufficiently enabled. The patent disclosed certain antibodies and methods for finding others, but the claims covered a much wider class than the disclosure adequately taught. The Court emphasised that the broader the scope claimed, the more the specification must enable. This principle is highly relevant when AI generates a large family of compounds but the patent describes only a limited number of them.

LII / Legal Information Institute

+1

2. In re Wands, 858 F.2d 731 (Fed. Cir. 1988)

The US Federal Circuit explained that enablement depends on whether the disclosure requires undue experimentation. Relevant considerations include the quantity of experimentation, guidance provided, examples, predictability of the field and the level of ordinary skill. This framework helps assess whether an AI-generated compound can be reproduced using the patent's disclosed instructions.

Justia Law

+1

3. O’Reilly v. Morse, 56 U.S. (15 How.) 62 (1854)

The US Supreme Court rejected an excessively broad claim covering every method of achieving a result when the inventor had not disclosed how to implement the full claimed scope. Applied to AI-generated compounds, a claim cannot validly monopolise an entire functional class merely because an algorithm can search for possible members of that class.

LII / Legal Information Institute

Significance

Patent enablement ensures that AI-assisted drug discovery contributes reproducible knowledge rather than speculative claims over vast numbers of untested compounds. Applicants should disclose enough technical information to support the actual scope of their claims, while carefully distinguishing predicted properties from experimentally verified results. In India, Section 10 provides the principal disclosure framework, while the cited US cases offer persuasive comparative guidance rather than binding Indian precedent.