Could AI Notetakers Put Your Patent Application at Risk?

In our recent article, Using AI to Prepare a Patent Filing: What Inventors (and Attorneys) Need to Know, we considered the potential risks of sharing confidential invention details with large language models (LLMs).

 

But what if that information is shared without the inventor realising it?

 

Imagine an inventor discussing a new technology with colleagues or investors over a video call. During the meeting, they explain how the invention works, what makes it different and the technical problems it solves.

 

Unbeknownst to the inventor, another participant has enabled an AI notetaker, which records, transcribes and processes the entire conversation.

 

Although the inventor has taken care not to disclose the invention publicly, a detailed account may now be stored on a third-party platform.

 

The question is, could this compromise the confidentiality of the invention?

 

Why confidentiality matters for patent applications

One of the key requirements for obtaining patent protection in the UK and Europe is novelty. This means that an invention must be new at the date the patent application is filed.

If information about an invention is made available to the public before filing, it may become part of the prior art and could prevent the invention from meeting the novelty requirement.

 

Importantly, a disclosure does not need to be made to a large audience to count as a public disclosure. Depending on the circumstances, sharing information with even one person who is not bound by a confidentiality agreement may be sufficient.

 

For this reason, inventors are generally advised to keep their inventions confidential until a patent application has been filed and to use appropriate confidentiality agreements, such as a non-disclosure agreement, when discussing their inventions with third parties.

 

Is using an AI notetaker the same as making a public disclosure?

Using an AI notetaker does not automatically make an invention publicly available.

 

The key issue is whether the invention’s details have been made accessible to anyone who is not bound by a confidentiality agreement.

 

When an AI notetaker records and processes a meeting, confidential information may be transferred to external service providers. If those providers are free to access or disclose the information without confidentiality restrictions, this could amount to a public disclosure and potentially jeopardise patent protection. As discussed in our previous article, the Upper Tribunal in UK v Secretary of State for the Home Department [2026] UKUT 81 (Hamid) raised serious concerns about uploading confidential information to open AI systems, while distinguishing closed systems that preserve confidentiality.

 

Although that case did not concern patents, it highlights that information shared with an AI service should not automatically be assumed to remain confidential.

 

Before using an AI notetaker to discuss an invention, inventors should understand how their meeting data will be used, stored and shared, and what confidentiality protections are in place. In particular, they should consider the following questions:

1. Is the meeting content used to train AI models?

Some AI providers may use customer information to train or improve their systems, while others expressly prohibit this.

 

Where confidential invention details are involved, it is important to establish whether the provider and any third-party AI processors are contractually prevented from using the information for training.

 

The danger of using meeting transcripts to train LLMs is that a model could potentially reproduce confidential details of an invention to third parties worldwide.

2. Who can access the recordings and transcripts?

The AI notetaker provider may use external transcription services, cloud infrastructure or AI model providers.

It is important to understand whether these third parties are subject to appropriate confidentiality obligations and whether any human review of meeting content is permitted.

 

3. How long is the information retained?

Some services store recordings and transcripts indefinitely unless deleted, while others allow users to set retention periods.

Even where information is not used for AI training, its continued storage may create additional opportunities for unintended disclosure.

 

4. Who can access the meeting notes?

AI-generated meeting notes and transcripts may be automatically shared with participants or made available to others within an organisation.

 

This means confidential details of an invention could reach people who were not part of the original meeting and who may not be bound by the same confidentiality agreements.

 

Takeaways

AI notetakers can be useful tools, but their convenience should not come at the expense of confidentiality. Where a meeting concerns an invention that has not yet been protected by a patent application, a cautious approach is advisable.

In particular:

• Check before the meeting whether any participant intends to use an AI recording or transcription service, and disable automatic notetakers unless approved.
• Review the provider’s terms to understand how meeting data is used, stored and shared, in particular check whether it is used to train AI models and who can access it.
• Check the sharing settings to ensure that recordings and transcripts are not accessible to unintended recipients.
• Consider traditional note-taking where the confidentiality arrangements of an AI service cannot be established.
  
  

Some AI notetakers offer safeguards designed to protect confidential information, but inventors should check the terms of the particular service rather than assuming these protections apply.