Alex Bajjon

May 2025

In order to obtain a patent, an invention must meet various criteria. One of the most important criteria is novelty, meaning that an invention must not be publicly disclosed before the patent application was filed. Disclosure may take many forms and includes disclosing in writing (e.g. in an earlier patent application or other written publication), orally (e.g. at a conference or to a group of friends in the pub!) or any other means, where the invention is told/shown to someone who is not under an obligation of confidence.  

Patent attorneys are consistently reminding their clients of the importance of filing a patent application before telling any other person about their invention.  However, in today’s world, it is not just telling other people about their invention that inventors need to be cautious of.  

Most commercial AI models, including widely used tools like ChatGPT, are cloud-based. When a user interacts with these models, the input data is typically transmitted to and processed on remote servers. While some platforms offer privacy guarantees, not all AI providers explicitly promise confidentiality or data isolation and, in fact, most free AI models do not offer this. This means that sharing the details of an invention with an AI chatbot could potentially be interpreted as a public disclosure, particularly if the AI provider retains logs, uses the data for training, or shares it with third parties. 

Although the question of whether disclosing an invention to an AI chatbot can constitute a novelty-destroying disclosure for the purpose of patentability has not yet been tested in the courts (at least, not to the author’s knowledge), most patent professionals see this as an extremely high risk. If an AI provider does not offer clear confidentiality terms and protection, then using their tools to research or describe a novel invention might prejudice any patent protection that is later pursued.  

With all of the opportunities that AI can offer in the R&D process, it is all too easy to involve AI in the development of a new invention, only to later discover that this now risks the patentability of the invention.  Steps that inventors or innovating companies can take to avoid being prejudiced by their use of AI when it comes to patent protection include: 

• Filing first and talking later – This has been the long-standing rule for patent protection.  As the best safeguard, a patent application should always be filed before discussing the invention with AI tools (as well as other humans!), especially those not designed with enterprise-level confidentiality safeguards. 

• Using confidential AI platforms – If the use of AI is crucial to the development of the invention, then (paid for) platforms that explicitly state your data won’t be stored or used for training should be chosen and the use of free AI models (such as ChatGPT), which do not have as robust privacy agreements, should be avoided. 

• Educating teams and colleagues – It is all too easy for a single member of a team to destroy the patentability of an invention by simply typing the wrong thing into an AI model.  Ensuring that all those involved in R&D (e.g. engineers and collaborators) understand the risks of casually using AI tools for invention development and searching is critical! 

• Consulting a patent attorney –  Of course, involving your patent attorney early in the innovation process to ensure proper protection strategies are in place is always worthwhile.  Don’t leave it too late and only involve your attorney once the damage has been done! 

Although AI tools are powerful allies in innovation, they must be used wisely.  Inventors may not view typing information into an AI bot the same as telling another person about an invention, but under current legal frameworks, inventors should treat AI disclosures with the same caution as any public blog post, press release, or academic publication. As always, the safest route is to protect your invention with a patent filing before you start chatting with your AI assistant!