Login | September 08, 2024
USPTO issues clarifying guidance on the use of AI
RICHARD WEINER
Technology for Lawyers
Published: June 28, 2024
The United States Patent and Trademark Office has issued some new guidance/ clarification on the use of artificial intelligence (or, I suppose, computer-generated human-adjacent creativity or something like that), based on how current rules cover AI use.
The guidance, published on April 11, ties up some loose ends that were left hanging by the agency and several court decisions, covered here before.
The guidance clarifies how existing USPTO rules apply to the use of AI to creations submitted to the office.
The guidance starts by emphasizing that the current rules cover the issue in large part, including the duty of candor/ good faith about the inventor/author, the signature requirement, client confidentiality, foreign filing regs and the office’s electronic filing policies.
Then the guidance goes into detail about the requirement of human verification.
Practitioners, it says, must double-check any work product generated by AI to assure relevancy and accuracy, particularly because of the propensity of legal AI to generate hallucinations in things like facts, case law, and case citations.
This is the warning that AI is just the start of legal research. Hopefully, all lawyers should know this by now.
The matter of checking art is also covered in basically the same way.
The guidance refers back to a February guidance on AI-assisted inventions to remind everyone that inventions in patent applications must have a “significant contribution by a human inventor.”
If not, that must be disclosed under current rules.
How about AI interacting with the USPTO?
Bet you didn’t think about that.
The guidance states that if anyone wants to use USPTO data to train AI, that data must be purchased rather than stolen (the latter of which is industry standard).
In addition, practitioners must prohibit AI tools from inputting practitioner signatures when using when using AI to fill out forms on the website.
Also, don’t let an AI create its own account. Yikes who thinks of this?
Finally, the guidance emphasizes that practitioners can’t violate the USPTO’s client confidentiality policy when using AI tools and that an AI tool won’t share confidential information.
So that’s that.
Trying to prevent some kind of future-AI patent apocalypse is a good thing.