The Company Playbook for AI-Assisted Inventions

Most of what has been written about AI and inventorship comes from law firms, and it is written for other lawyers. That is useful if you are prosecuting a patent. It is less useful if you run an engineering organization and need to know what your team should do differently on Monday morning.

This post is the operating view. It covers what changed, what did not, and the three habits we see companies adopt when AI tools are part of daily engineering work. It is not legal advice, and your patent counsel should shape the final version of any policy you adopt.

What the USPTO actually said

In November 2025 the USPTO published revised inventorship guidance for AI-assisted inventions and rescinded its February 2024 guidance in its entirety. The core message of the new guidance is short: the same inventorship standard applies to every invention, whether or not AI was used along the way. Only natural persons can be named as inventors, and the agency describes AI systems as tools that people use. You can read the notice in the Federal Register.

For an operator, the takeaway is not that the rules got easier or harder. It is that the question hasn't changed. Who conceived the invention? Conception has always been the touchstone, and the guidance points back to it. What changes in practice is how hard it can be to show conception later, when a design emerged from a long session with a coding assistant and nobody wrote down which ideas came from whom.

Why this is an engineering problem, not just a legal one

Counsel can only work with the record your team leaves behind. If the record is a merged pull request and a chat history that got deleted, the conversation about inventorship starts from memory. Memory is a poor witness, especially eighteen months later during diligence.

The companies that handle this well treat it as process design. They do not ask engineers to become patent experts. They add a few lightweight steps to work that is already happening.

Habit one: capture the human contribution while it is fresh

The most useful record is a short note, written close to the work, that answers a few questions. What problem were you solving? What approach did you decide on, and why? Where did a tool suggest something, and what did you change, reject, or combine? That last question is the one most teams skip, and it is the one that matters most now.

We go deeper on what that note should contain in how to document human conception when engineers use AI tools.

Habit two: update the invention disclosure form

Most disclosure forms were written before anyone pasted a design question into a chat window. They ask for a title, a description, and the names of contributors. They rarely ask how the invention was developed or what tools were involved. Adding a handful of fields closes most of the gap. We walk through them field by field in what to put in your invention disclosure form.

Habit three: decide where unfiled inventions are allowed to go

This is the issue that tends to surprise leadership. An engineer working on an unfiled invention pastes the core of it into a consumer AI tool to get help with a description or a diagram. Depending on the tool's terms and settings, that content may be retained or used in ways the company did not intend. Whether that creates a disclosure or trade secret problem is a question for counsel, and the answer depends on the facts. The safer move is to decide the rules in advance. We cover the options in the confidentiality problem with unfiled inventions and AI tools.

A one-page policy is enough to start

For most mid-market teams, a first version covers four points:

• Which AI tools are approved for work involving confidential or unfiled technical material
• A short contribution note for any work that might become an invention
• An updated disclosure form that asks how the invention was developed
• A named person who routes possible inventions to patent counsel for review

None of this slows shipping. It turns a question that usually gets answered in a hurry, during a financing or an acquisition, into one that gets answered as the work happens.

Where we fit

At Idea Partnership we help companies surface and develop inventions and prepare the technical record that independent patent counsel uses to assess and file. If your team is building quickly with AI in the loop and nobody owns this process yet, a short call is a good place to start.