Summary: The Indian Patent Office will be using artificial intelligence in the patent examination process. Guidelines for such use have been issued recently, outlining the terms of use, safeguards, and prohibited activities.
The Office of the Controller General of Patents, Designs and Trade Marks (CGPDTM) has issued Guidelines for the Use of Artificial Intelligence in Patent Examination Procedures (Guidelines), available here. The Guidelines acknowledge that, going forward, artificial intelligence (AI) will be a part of the patent examination process at the Indian Patent Office (IPO).
1. Objective
The Guidelines seek to lay out the use of AI in patent examination in a manner that supports efficiency and quality, while preserving confidentiality, accountability, consistency and the independent application of mind by the Examiner or Controller.
The Guidelines clarify that the Examiner or Controller remains responsible for the official act even where AI has been used. Any AI-assisted output that influences examination must be manually reviewed and validated, and can be adopted only after the officer is satisfied as to its correctness, relevance and appropriateness.
The fundamental policy position may be summarised as:
“Artificial Intelligence is intended to assist and not replace the functions or the works of the Examiner or Controller.”
2. Uses of AI in Patent Examination
The Guidelines list out the activities for which AI may be used by officials, while highlighting the level of human intervention in each case. The Guidelines envisage AI mainly as a search, screening, organisation and drafting-assistance tool across different stages of examination, rather than as the decision-maker.
Key activities are summarised below:
| S.No | Activity | Role of AI | Human involvement |
| a. | Patent classification
|
Generate IPC/CPC classifications relevant to the invention | Final classification must be manually verified and appropriately selected. |
| b. | Prior-art searching | Generate search terms, synonyms, related expressions and concept clusters. | Final search terms must be manually selected, and search strategy must be refined. |
| Prior art search using officially subscribed AI tools. | Query must be framed, and if required, refined, manually. Each result must be verified manually and the final result should be selected based on independent judgement. | ||
| c. | Claim feature extraction | Break down complex claims into structured feature lists for comparison. | Features, relationships and ranges must be manually checked before being relied upon for analysis. |
| d. | Translation support | Obtaining a preliminary understanding of foreign-language patent documents. | Technical meaning must be verified by reading the translated text along with drawings, figures, and surrounding disclosure. |
| e. | Prior-art mapping | Preliminary claim mapping and structured comparison with prior art. | Each feature must be independently verified |
| f. | Novelty/inventive step | Preliminary assistance in claims mapping, structured comparison, and internal organisation of analysis | Final conclusions on novelty and inventive step must be based on manual analysis, after independently verifying any AI-generated inputs. |
| g. | Sufficiency and clarity | Identify possible ambiguity, indefiniteness, drafting defects and possible gaps in enablement or fair basis. | Manually review of complete specification and relevant common general knowledge, and independently assess sufficiency and clarity requirements. |
| h. | Drafting | Improve the grammar, structure, coherence, headings and readability of an office communication. | Manually formulate and control the substantive reasoning. AI may help refine an existing draft, but it cannot replace independent human reasoning.
|
| i. | Retrieval of legal and technical information and case-law understanding | Identify potentially relevant case law, scientific references, patent citations and technical literature as well as assistance in summarising judgments or explaining | Each source must be independently verified. AI may be used only for preliminary understanding, and the relevance and applicability of the identified material must be independent assessed. |
3. Prohibited Uses
The Guidelines also list out uses of AI that are not permitted:
- Unpublished patent material
Officials may not enter unpublished patent application contents, confidential office records or internal deliberative material into public AI tools. - Substantive matters
AI cannot be the substitute for human application of mind on substantive examination matters, including novelty, inventive step, industrial applicability, sufficiency, clarity, or unity of invention. - Official communications
Office actions, first examination reports, hearing notices, decisions or other official communications may not be issued solely on the basis of AI-generated output without adequate human oversight. - Unverified citations
AI-generated case law, prior art, scientific literature or other references cannot be relied upon without independent verification from an authentic source. - Unreviewed AI-generated content in official communications
AI-generated material may not be used in official communication without review, correction, and adoption by the concerned officer. - Decision-making
AI may not be used for decision-making affecting rights of applicants/patentees or third parties, especially in arriving at conclusions in cases where situations involve opposing parties, for instance, opposition proceedings.
4. Safeguards and Administrative Measures
The Guidelines contemplate an institutional framework for governing the use of AI, including the following components:
- Recording of AI use
The use of AI may be recorded, including: the AI tool used; nature of use; date of use; and other information needed for supervision, audit or quality review. Appropriate disclosures may be made available to stakeholders to improve the confidence in the examination process.Annexure 2 of the Guidelines contains a checklist and the template of a declaration as to the use of AI to be submitted by officials. - AI Governance Committee
A dedicated AI Governance Committee is proposed within the CGPDTM, including nominated representatives from the Examination Division, IT Office and QMS Division, with representation covering different technology groups.This Committee would be responsible for:- reviewing and approving AI tools;
- categorising uses as permitted or prohibited;
- prescribing safeguards and approval conditions;
- conducting pilot studies;
- reviewing quality issues, errors and complaints;
- periodically revising the Guidelines; and
- consulting stakeholders and AI experts.
- Training and capacity building
Training and capacity building of officials will cover:- AI functionalities and limitations;
- confidentiality and prompt discipline;
- verification of AI-assisted outputs;
- responsible use of AI in public service;
- identification of hallucinations, unsupported claims and fabricated citations; and
- permitted and prohibited uses.
- Audits and incident reporting
Independent audits, impact assessments, feedback mechanisms, and incident- reporting systems will also be a part of the process.
5. Key takeaways for Patent Applicants and Practitioners
The Guidelines, although directed at the use of AI in patent examination, will impact patent prosecution in more ways than one.
First, applicants should address actual disclosure of cited prior-art documents, rather than merely responding to an AI-generated mapping. Since AI will be used to identify similarities between a claim and prior art, each claimed element and the relationship between various elements should be independently verified. Mere semantic or functional similarities identified by AI should not be assumed to establish disclosure of a claimed feature.
Also, practitioners should keep in mind the likely use of AI in examination at the drafting stage itself. Specifications as drafted should avoid gaps in clarity, support and sufficiency of disclosure. The scope and technical meaning of the invention should not be susceptible to misinterpretation during AI-assisted examination. Therefore, technical relationships between features, numerical limitations, functional dependencies and other claim-defining aspects should be expressly and consistently disclosed.
Conclusion
The Guidelines signal a formal move towards AI-assisted patent examination. While the IPO recognises the potential of AI in the patent examination process, it also recognises the limitations of AI (hallucination, misreading technical language, overlooking claim limitations, generating false pattern matches, and fabricated / inaccurate citations). With these Guidelines, the IPO seeks a balanced approach, where the future is not where “AI examines patents”, but where “Examiners examine patents with the assistance of AI.”
Stakeholders must, on their part, recognise that AI-assisted examination will make careful drafting, precise claim language and comprehensive disclosure increasingly important from the very start of the patenting process.
English
Chinese (Simplified)
Japanese