A well known problem of generative Artificial Intelligence is the tendency to “hallucinate” and produce statements which look factual but are not. Like other sectors, legal professionals are rapidly deploying AI, and so must take care to use it responsibly.  

In this context, the United States Patent and Trademark Office (USPTO) recently announced guidance for practitioners and the public for navigating the complexities of AI in patent and trademark proceedings. This guidance aims to address the growing integration of AI into the preparation and prosecution of patent applications as well as other IP filings, and recognises the immense potential and inherent risks of these technologies. 

With appropriate privacy and confidentiality protections in place, AI writing assistant technologies can be useful tools in IP-related drafting, offering capabilities that significantly enhance efficiency and accuracy in tasks such as legal research, prior art searches, and document preparation. For instance, AI can analyse vast datasets faster than human counterparts during semantic searches of patent databases, potentially identifying patterns and connections that might go unnoticed otherwise. As with human drafters, these tools pose risks, such as generating incomplete or inaccurate outputs and introducing bias or errors into legal documents.  


The Guidance 

The USPTO’s guidance therefore seeks to clarify the rules and responsibilities concerning the use of AI in IP practice, such as for patent attorneys, trade mark attorneys, agents, and law firms.

Despite the convenience of AI-generated content, the guidance requests a human in the loop.  Responsible parties are asked to verify AI-generated outputs, thereby aiding that submissions to the USPTO are accurate. For example, it is a common best practise for regulated IP firms to ensure a regulated person is always in the loop (such as a qualified patent attorney) and then seek verification from the subject matter expert (commonly, the inventor or applicant) that the contents included in the filing are accurate and factual prior to filing.  

Practitioners are further reminded by the USPTO that they must ensure that the use of AI does not lead to unauthorized disclosure of client information, adhering to strict confidentiality obligations. 

The guidance does not prohibit the use of AI in the IP realm for drafting, but stresses the importance of using these tools responsibly. The USPTO emphasises it commits to ongoing discussions with stakeholders through initiatives like the AI and Emerging Technologies Partnership. This dialogue aims to continually assess and update guidelines as necessary to keep pace with technological advancements.

These kind of reminders are all good common sense.

However, there is something else interesting in the guidance. First, the USPTO reminds practitioners of their obligation to the USPTO to act with candor and good faith under 37 CFR 11.303. This duty requires disclosure of all information known to be material to patentability.  This is common practice already when it comes to “prior art” documents that are known about at the time of filing, or become known during the life of an application, so that the Examiner can get the complete picture to make an informed patentability assessment.

However, the new guidance also states:


Further, practitioners and others involved in a matter before the USPTO may be required to disclose certain known facts to the USPTO under their duty of candor and good faith. For example, in patents and patent applications, all patent claims must have a significant contribution by a human inventor. Thus, if an AI system is used to draft patent claims that are submitted for examination, but an individual [involved with the application] has knowledge that one or more of the claims did not have a significant contribution by a human inventor, that information must be disclosed to the USPTO.

What this means in practice will become clear over time.  In our view, it would be simply negligent to just “chuck in” some purely AI-generated patent claims to a patent office without full human review.  Even if an AI tool was used for this sensitive job, the claims are the key part of a patent which are drafted with utmost care.  For this reason, we expect it to be very rare that this kind of submission would need to be made. But the fact that the USPTO have had to put out this clarification shows just how concerned they are to maintain quality. 

Implications and Boundaries

In general, practitioners will need to be vigilant to AI changes and ensure that their use of AI aligns with legal standards and ethical practices of the territory they are operating and filing within.   

The USPTO’s guidance serves as a cautionary framework for integrating AI tools into their practice responsibly. It necessitates a balance between leveraging AI for improving the quality of legal work while also maintaining standards of professional responsibility.

European practitioners in particular will already be familiar with the EPO’s own notoriously high standards for sufficiency and the standards of what is required for an enabling disclose, the use of AI assistance must be used to enhance these aspects, not be detrimental to them.  

Practitioners will need to be observant to where the boundaries lie in this rapidly evolving space as they navigate the reality that AI tools are increasingly embedded at all stages of research and development.  


At Scintilla, we help innovative companies get a grip on their intellectual property. Our unique commercial approach combines registration of patents and trade marks with strategic input so that IP can be a springboard for business growth. If you would like to discuss your IP needs, do contact us or book a free initial consultation!