India's Patent Office Issues Guidelines on the Use of Artificial Intelligence in Patent Examination

House no. 316, Ground Floor "A" Sector, Yelahanka New Town Bengaluru- 560 064

1203, The Spire-2, Shital Park BRTS, 150 Feet Ring Road,Rajkot - 360007 Phone : +91 281 242 731

203, 2nd Floor, Shriji Palace,Savsar Plot, Main Road, Morbi-363641 INDIA Phone:91 2822225263

312-313, 3rd Floor,
Abhishek Complex, Akshar Chowk, Old Padra Road,
Vadodara 390 020 INDIA
Phone: (0265) 2322015

The Office of the Controller General of Patents, Designs and Trade Marks (CGPDTM) has released the “Guidelines for the Use of Artificial Intelligence in Patent Examination Procedures.” The Guidelines set out how Examiners and Controllers may use public and private AI tools to assist patent search, classification, translation, drafting and analysis — while keeping the statutory and quasi-judicial functions of the Patent Office firmly in human hands. Below is a summary of the key points for stakeholders.

1. Objective & Scope
The Guidelines aim to regulate AI use in patent examinations in a manner that improves efficiency and quality while preserving confidentiality, accountability, consistency and the independent application of mind by the Examiner or Controller. They apply to AI use in screening, classification, search, translation support, drafting support, technical comparison and knowledge retrieval.

2. Key Concepts

• Public AI tools – free/commercial consumer-facing platforms trained on broad public data; versatile but of limited depth in specialised domains.
• Private AI tools – closed, access-restricted systems (including those built into the Office's subscribed search databases) suited to sensitive or proprietary data keeping the system closed with heightened security measures to safeguard against unauthorized access.
• AI-assisted output – content generated collaboratively with a “human in the loop” providing direction, context and oversight.
• Prompt engineering & prompt discipline – careful, structured framing of prompts materially affects the relevance and quality of AI output.

3. Limitations & Risks Flagged
The Guidelines list below recurring failure modes that require human vigilance, including:

• Hallucination/fabrication – confident but unsupported facts, citations or case law.
• Omission of critical claim detail, amendments, disclaimers or qualifiers.
• False pattern matching – surface/keyword similarity mistaken for technical or legal relevance.
• Classification and search drift; over-breadth or over-narrowness.
• Black-box opacity, profile sensitivity and prompt sensitivity in outputs.
• Confidentiality risks from sharing unpublished material with external tools.
• Incomplete/outdated training data and bias in training data.
• Difficulty handling complex patent language, drawings and jurisdiction-specific legal nuance.

The overarching principle: AI is intended to assist, and not replace, the functions or the works of the Examiner or Controller. Every AI-assisted output must be independently reviewed, verified and adopted only after the officer is personally satisfied of its correctness — responsibility for the official act is never diluted or transferred.

5. Prohibited Uses
The Guidelines expressly bar:

• Entering unpublished application content, confidential records or internal deliberative material into public AI tools.
• Using AI as a substitute for the Examiner's/Controller's independent judgment on novelty, inventive step, industrial applicability, sufficiency, clarity or unity of invention.
• Issuing office actions, FERs, hearing notices or decisions based solely on AI output without adequate human oversight.
• Citing AI-suggested case law, prior art or literature without independent verification from authentic sources.
• Using AI-generated content in official communications without review, correction and adoption by the officer.
• Relying solely on AI for decisions affecting applicants'/patentees' or third-party rights, especially in contested proceedings such as oppositions.

6. Administrative Measures

• Recording of material AI use (tool name, nature and date of use) may be prescribed for supervision, audit and quality review.
• An AI Governance Committee (drawn from Examination, IT and QMS divisions) will review and approve tools, categorise permitted/prohibited uses, prescribe safeguards, run pilot studies and periodically revise the Guidelines.
• Training and capacity building for officers on AI functionalities, limitations, confidentiality risks, prompt discipline and identification of hallucinations.
• Independent audit, impact assessment, feedback mechanisms and incident reporting may be introduced.

An accompanying Checklist & Declaration (Annexure-II) requires officers to confirm, among other things, that AI use was permitted, no confidential material was entered into unapproved tools, all citations and technical/legal statements were independently verified, and that the final analysis remains the officer's own application of mind.

Applicants and practitioners can expect AI to play a growing but tightly supervised role in Indian patent prosecution — aiding classification, search, translation and drafting efficiency — while novelty, inventive step and other substantive determinations remain firmly the product of independent examiner judgment, backed by verified sources. Practitioners should anticipate continued human-reasoned office actions, with any AI assistance disclosed and auditable where required.
IP India has added Annexure-I (illustrative examples of the use of Artificial Intelligence in patent examination) and Annexure-II (checklist and declaration) on its website, providing useful reference material for stakeholders to review while navigating AI-related patent applications.