Who Owns the Code Your AI Wrote?
The code your team generated with AI last week might not legally be yours.
Here is why.
Copyright protects human authorship. That is the whole foundation. US courts and the Copyright Office have been consistent on this. Type a prompt, get an output, and that output is not human created. AI is treated as a tool, like a camera. Only the parts a human creatively shaped are protected.
This is not new logic. It goes back to an 1880s Supreme Court case over a photograph of Oscar Wilde. Ordinary snapshots were ruled not protectable. The photo earned copyright only because the photographer posed the subject, arranged the light, and composed the shot. Human creative input was the line.
Now apply that to AI-generated assets. Code, text, images, voice, video. If the human contribution was ‘I wrote a prompt,’ a court may call the result ordinary. Ordinary means uncopyrightable. Uncopyrightable means a competitor can copy it, ship it, and you have no claim.
For a product company, that should land hard. The firmware you assumed was your IP may already be sitting in the public domain.
The fix is not to stop using AI.
It is to keep a real human creative component, document exactly what the human did versus the AI, and remember that copyright was never your only moat. Trade secrets, proprietary data, and contracts still protect you. Sometimes better.
So here is the question every engineering leader using AI should sit with. How much of what you shipped this year would survive a lawsuit?
The full IEEE Computer piece this is based on (Aleix Martinez, June 2026) lays out the legal reasoning and the practical steps to protect your work.
Read the IEEE Computer article
