You design a diagnostic tool with three key elements: a biomarker panel you identified, a data collection protocol you developed, a risk threshold you optimized through testing.

Then you feed it to Claude: “What’s missing?” Claude suggests a fourth element: a normalization step that makes it more robust. Can you patent this four-element system?

Inventorship: What You Actually Did

Under patent law, only human beings can be inventors. An inventor who uses AI to test their idea, research a hunch, or explore embodiments is still an inventor. An inventor who uses AI to suggest elements and blindly adopts them is not an inventor for those elements.

The distinction is simple: did you make the invention, or did you just adopt it?

Scenario A: You Validated It. You tested the normalization step. You understood why it works. You modified it to fit your protocol. You documented the data. You significantly contributed. The patent holds.

Scenario B: You Adopted It Blindly. Claude suggested it. You copied it without testing, without understanding the mechanism, without technical judgment. You didn’t invent the fourth element. You adopted it. Inventorship is weak.

The Honest Conversation

In that sense, the duty of candor renders AI into something much less new and exciting: another opportunity for an inventor to be frank about where ideas came from. You’ve always owed the USPTO honesty about whether you conceived an element, collaborated on it, built on prior work, or adopted an external suggestion. AI doesn’t change that obligation—it just makes it more visible.

The same way you would disclose that a colleague suggested a design approach, or that you adapted a published technique, you now disclose how you used an AI tool. The tool matters less than the truthfulness. What matters is knowing, before you file, whether you actually invented what you’re claiming to have invented.

 Before You File Ask yourself:

 Did I understand why this element works?

• Did I test it?
• Did I modify it based on my judgment?
• Or did I just accept what AI suggested?

It is harder than it sounds. It requires you to slow down. To test Claude’s suggestions instead of adopting them. To modify and improve instead of copy-pasting. To understand why something works before claiming you invented it. But that’s what invention actually is.

The USPTO requires “significant human contribution.” But that’s legal language for something simpler:  did you make deliberate choices about this invention? If you can’t answer yes, you didn’t significantly contribute. Your inventorship claim is weak—not because of a rule, but because you didn’t actually invent it. That’s what you need to know before you file. Just whether you can look at each element and honestly say: “I invented this.”