Only real people can patent inventions – not AI, US Government says
New US Patent and Trademark Office guidance confirms that only human beings can be listed as inventors, even when AI tools play a major role, and that a “significant” human contribution is required for a valid patent. Commenters debate whether this is a sensible way to preserve the intent of the patent system or an unenforceable formality, given that people can simply omit mentioning AI and still claim credit. The broader exchange questions how AI-generated ideas should interact with concepts like prior art, ownership, and authorship, and whether current IP frameworks can cope as generative models scale.
AI as Prior Art and “Poisoning the Well”
- Several comments propose using LLMs to mass‑generate inventions and publish them to block future patents (a “tome of prior art”).
- Others argue AI output may be noisy, unworkable, or implausible, so its value as prior art is limited until humans validate and filter it.
- There is uncertainty whether AI‑generated text will legally count as prior art; some expect future rulings could exclude it, enabling “laundering” AI ideas through human applicants.
Purpose and Coherence of the Patent System
- One camp says patents exist to reward R&D investment and encourage disclosure instead of trade secrets.
- Another argues the real driver is belief in “special” human creation and ownership of ideas; if machines can create cheaply at scale, that moral basis erodes.
- Several comments say the system is already “broken” or “barely useful,” with large corporations spamming marginal patents and blocking downstream innovation.
USPTO Ruling: Human Inventor and “Significant Contribution”
- The new guidance: AI can assist, but a patent’s inventor must be human and must have made a “significant contribution.”
- Some see this as sensible: AI is just a tool like CAD; humans still prompt, refine, test, and prototype.
- Others think the “significant contribution” standard is easy to bypass and mainly symbolic or political theater.
- Clarified distinction: corporations can apply for patents but cannot be listed as inventors.
Enforcement, Obviousness, and Honesty
- Commenters note it’s hard to prove whether AI was used; enforcement mainly arises when patents are litigated.
- If an invention is easily produced by a public LLM, some argue it should fail the “non‑obviousness” test.
- Courts, opposing parties, and prior‑art challenges are seen as the main practical checks.
AI Personhood and Rights
- Many reject giving current AIs inventor or author status; they lack legal personhood, responsibility, and economic motivation.
- Others anticipate long‑term debates over AI or AGI rights, analogizing to corporations, animals, or hypothetical aliens, and warn about ethical and political complexity.
Side Threads: Copyright, Recipes, and Corporate Power
- Discussion branches into copyright vs patents (e.g., recipes generally not copyrightable, but industrial food processes patentable).
- Some see the ruling as a partial check on AI vendors claiming ownership of users’ AI‑assisted work.
- There is recurring skepticism about corporate personhood and its interaction with future AI entities.