Before you prompt: AI and the risk of public prior disclosure for patentable inventions

For patents, prior disclosure of an invention can have consequences for patentability and may render it impossible to obtain a patent in almost every jurisdiction in the world.

The recent proliferation of artificial intelligence (AI) tools, particularly large language models (LLMs), has potentially introduced a new avenue for prior disclosures. This article examines the legal implications of using AI tools in connection with inventions and highlights the risks they pose to inventors and patent applicants.

Novelty and obviousness – Prior art

An invention must be novel and non-obvious to be patentable. In Canada, this means the subject matter of a claim must not have been previously disclosed in a way that made it available to the public anywhere in the world, as provided by the Patent Act.

An invention’s lack of novelty is established through anticipation. Anticipation is assessed on a claim-by-claim basis by asking whether the prior disclosure, when understood by a person skilled in the art in light of their common general knowledge, provides both a description of the claimed invention (disclosure) and sufficient instructions to enable the invention to be practised (enablement).

In addition, an invention must be non-obvious. Prior disclosure can be used to establish that an invention would have been obvious, which can preclude or invalidate a claim. Some examples of prior disclosure include publication in an academic paper or journal article, a presentation at a symposium or conference, a publicly accessible website post or other online publication, or even an unguarded conversation in a public place.

Similar approaches to novelty (anticipation) and obviousness (inventiveness) are found in virtually all patent systems worldwide.

How AI tools may create unintended disclosures

AI providers typically address how a user’s prompts are handled through their terms of service (user input data). For example, the terms of service may:

• grant the provider rights to use, reproduce, or publish user input data for any purpose, including system improvements;
• allow users to opt out of user input data collection or use;
• permit the provider to use user input data for specified purposes even when a user has opted out;
• state that the AI provider will de-identify user input data before using it for AI model training purposes; or
• apply different user input data collection rules to free versus paid versions of the tool.

Whether such a disclosure constitutes anticipation or prior art remains uncertain and likely depends on the input and the AI system’s data handling. The legal standard for what constitutes a “public” disclosure in Canada does not require proof that anyone actually accessed the information. The test is objective: whether the public had a real opportunity to access it. In particular, the Federal Court of Appeal explained that an anticipating description, such as in a book, will invalidate a patent if the book is on a shelf of a library open to the public, whether or not anybody read the book, and whether or not it was situated in a dark and dusty corner of the library. In other words, if the information is available to the public, then the public has the right to make and use the information in the book without hindrance from a monopoly granted.

The U.S. Patent and Trademark Office has published a Guidance on Use of Artificial Intelligence-Based Tools in Practice Before the United States Patent and Trademark Office cautioning that “use of AI systems to perform prior art searches, application drafting, etc. may result in the inadvertent disclosure of client-sensitive or confidential information to third parties through the owners of these systems, causing harms to the client.”

However, neither the Canadian Courts nor the Canadian Intellectual Property Office have directly addressed the prior disclosure and AI systems. It therefore remains uncertain whether submitting invention details to a commercial AI tool would constitute anticipation in Canada.

Practical recommendations

Until clear regulatory guidance emerges in Canada, inventors and patent applicants should consider the following disclosure-related considerations when using AI tools:

• Understand the AI provider’s terms of service and opt-in or opt-out mechanisms, including how input data is stored and used.
• Exercise caution when using AI tools in connection with invention details or patent work.
• If an inadvertent disclosure has already occurred, consider filing a patent application promptly to take advantage of Canada’s 12-month grace period, which exempts an inventor’s own public disclosure if an application is filed within that timeframe.

For questions about AI and your invention, or for assistance with patent applications, please contact DLA Piper’s Canadian Intellectual Property & Technology Group.