Introduction
The Indian Patent Act, 1970, lays the foundational framework for Patents in India, this framework incorporates the process and the requisites for obtaining the patent registration by an applicant. Patent examination is a vital step in the entire process, it is through this step that the eligibility of an invention to obtain a patent registration is determined.
With the advancement of Artificial Intelligence technology and its ever widening uses, the Office of Controller General of Patents, Designs & Trade Marks published the Guidelines for the Use of Artificial Intelligence in Patent Examination Procedures (“Indian Patent Office AI guidelines”) on August 7, 2026[1]. These guidelines lay the foundation for the examiners to efficiently use AI tools in the examination process to boost the output without compromising the standards of examination of patent applications. Furthermore, striking the right balance of AI usage and human oversight and accountability, with these guidelines is of utmost importance because of the confidential and secure nature of the patent applications.
Objective
These patent examination guidelines are in furtherance of various policy documents issued by the Government of India that emphasize upon and supplement the idea that AI technology should be incorporated and utilised in a structured and responsible manner in public functions. These documents recognise its utility in improving efficiency and public service delivery, while emphasising human oversight, accountability, safety, transparency, privacy, risk-based governance, and institutional safeguards.
While these guidelines lay the affirmative framework for incorporation of AI technology in public functions, these guidelines also provide cautionary aspects to the unmindful use of AI technology and also highlights the importance of human oversight.
These guidelines, while defining the difference between public AI tools and private AI tools have also suggested the use of private AI tools for the proposed objectives based on the reasoning that private AI tools operate on closed environments and are trained using subject specific datasets and furthermore, these tools operate on closed corporate clouds which increases the security keeping in tune with the highly confidential nature of patent applications.
Limitations
These guidelines provide a strong point in favour of supervised AI usage, and have encapsulated several predictable issues which may arise if the AI tools are used without human oversight. However, it is imperative to understand that while this list is elaborate, it is not exhaustive and since this technology is rapidly advancing, new unforeseen challenges or issues may arise with the advancement of the technology and thus human oversight and vigilance is absolutely necessary.
The guidelines have identified and laid caution against issues like source of information and the subsequent output, i.e. AI tools are always learning, this learning includes the past interactions and user specific contextual patterns which may result in varying responses of same queries to different users, furthermore, the output is also based on the input prompts and therefore, careful drafting of prompts is necessary for a obtaining the intended output from the AI tools. Despite this, it is important to keep in mind that careful usage of prompts and output generated may include fabricated facts to fill gaps in the information. The authenticity of AI citations has been floating around, among controversial topics for some time now, where even High Court judges have spoken on the matter. Thus it is important to verify each and every information provided by the AI tools.
The issues with the AI tool’s capability to accurately understand the complete and complex context of the situation and the required output may become a grave challenge, while AI tools may identify surface level similarities, the necessity to understand the complex legal or technical context might be overlooked resulting in a research which is relevant as per pattern recognition but the context of the subject matter is completely or partly missing.
In addition to several listed risks of unsupervised AI usage, confidentiality is of utmost importance, any breach of confidentiality shall have dire consequences for the applicants due to unauthorised sharing or leakage of unpublished or internal material. Further patent applications are written in highly technical and legal language, which the AI tools may not be able to understand properly and therefore, the results might not be correct.
Uses
The patent examination guidelines while providing caution to the unsupervised use of AI and its associated risks, has also identified the use-cases for the AI tools and how they can increase efficiency and cut short the time for tasks such as generating prima facie IPC or CPC classifications, which may be authenticated or verified with the claims read along with complete specifications.
These guidelines also suggest that the AI may be used to generate search terms and concept clusters, as it will assist in suggesting related expressions, synonyms, alternative spellings, connected technical concepts and broader terms.
However, the guideline strictly emphasize that the examiner should use their discretion regarding the search terms liberally, to be able to differentiate which terms reflect core claims and which do not, further, they must also assess at their sole discretion which of the broad options are actually helpful for the search against the ones which won’t yield helpful results.
Use of AI for translation support is another use case scenario suggested by the guidelines, AI tools could be used for translation of foreign language documents for indicative understanding. However, the translation should be read together with the drawings and the related disclosure for a more comprehensive understanding
Furthermore, using AI tools for internal communication and early-stage work is a suggested use, however, the guidelines caution and suggest that AI tools for internal communications must only be used after the core of the intended communication message has been drafted by the officers and once it has been polished by the AI tools the communication should be checked to ascertain that the message still contains its exact contextual interpretation.
Prohibited Uses
These Guidelines have provided prohibitions on certain uses of AI. According to the guidelines feeding confidential data like unpublished application contents or internal office records in the public AI tools, furthermore, considering AI in a substitute role for the examiner’s or controller’s own cognitive thinking and diligence, instead of as a supplementary role to the examiner or controller has be enlisted as a prohibited use.
Furthermore, relying solely on the output of the AI tools without human oversight and verification for tasks like issuing office actions, FERs, hearing notices, decisions or other official communications, providing citations regarding prior art, scientific literature and other references have been also prohibited.
It has also been emphasised by the guidelines that AI tools should not be used as conclusive decision makers, AI tools should be utilised to help the examiner or controller in tedious tasks to increase efficiency. The thought behind integrating AI in the patent examination process is to provide supplementary help to the examiners or controllers.
Administrative Measures
The patent examination guidelines provide for the administrative measures that will be taken for the implementation of the guidelines. These measure involve publication of detailed out material use of AI tools in specific functions by the competent authority. Furthermore, a committee shall be convened to monitor the approval of the tools for usage, proper classification of allowed and prohibited uses of the AI tools, mandating the safeguards, overseeing the quality control and managing that an open dialogue is maintained with the relevant stakeholders and AI experts to stay ahead of any risks or contingencies that may arise with the advancement of AI technology.
The guidelines provide the composition of the AI Governance Committee, which shall consist of examiner(s) and controller(s) from the Examination Division of the Patent Office, officer(s) from the IT Office and from the QMS Division. The selection will be done on nomination basis to ensure that vide variety of expertise is maintained in the Committee.
Training and learning mechanisms to enhance the understanding of the officers in relation to function and limitations of the AI tools, proper usage of prompts and deeper understanding of the risks associated with confidentiality while using AI tools, proper verification of facts and clear understanding of allowed and prohibited uses, will also be initiated by the competent authority. In furtherance of the learning mechanisms the competent authority shall ensure independent audits, impact assessment, feedback mechanisms, and incident reporting systems.
Practical Examples and Accountability
The patent examination guidelines have through numerous illustrations using actual patent examination functions, highlighted the indispensable need of human oversight and verification. To substantiate the laid down principles, the guidelines have used hypothetical and illustrative conditions in a wide range of operations like novelty assessments, disclosure sufficiency, clarity objections, translation, and legal research
Furthermore, the guidelines have provided for safeguards in examination workflow through a checklist and declaration in Annexure-II of the guidelines. The checklist mandates that the officer complies with the office guidelines issued in respect of AI usage, is cautious while dealing with confidential information so as to prevent unintended disclosure, furthermore, the officer must verify all the AI output to maintain human oversight.
In furtherance of the above mentioned mandates, the guidelines also mandate that the officers properly report the exact function of the AI tool if it is utilised in any operational capacity. Thus, the declaration acts as a surety of personal accountability of the officers.
Conclusion
These guidelines were published on 7th August, 2026, by the Office of Controller General of Patents, Designs & Trade Marks without conducting a public consultation and feedback exercise, since these Guidelines affect a large number of applicants and other stakeholders including but not limited to legal practitioners, international patent offices, research institutions, and the broader public.
Since, the guidelines suggest adoption of AI tools and a significant change in the existing framework, the feedback from the larger section of people to be affected by the changes, should have been considered as it would have provided many helpful insights.
As the saying goes, four eyes see more than two, allowing for public feedback on these guidelines would have helped the find out any short comings of the guidelines and would have minimised the chances of any future discrepancies. In conclusion, public feedback would have strengthened the framework.
Author: Akshi Seem, Associate Partner and Anushrut Singh Rajawat, Associate
- News Details, Office of Controller General of Patents, Designs & Trade Marks at https://www.ipindia.gov.in/dynamic/news-details/119, Last accessed on 12/08/2026.
- NITI Aayog, Government of India, RESPONSIBLE AI #AIFORALL, Approach Document for India: Part-1 & Part-2
