This week, an artificial intelligence did something inventors have always done by hand: it discovered a better way to make a drug. As Dr. Alex Wissner-Gross reported in the June 17 edition of The Innermost Loop, OpenAI and the chemistry startup Molecule.one turned a frontier model loose in a robotic lab, where across 10,080 experiments the system picked its own research target, proposed approaches, ran the reactions, and taught itself a trick that lifted the yield of a stubborn synthesis used in dozens of FDA-approved drugs. That is a scientific milestone. It is also a legal problem, because AI-generated inventions raise a question that U.S. patent law has only just answered: who, exactly, is the inventor?
What You’ll Learn
What the AI Chemistry Breakthrough Means for Business
For years, companies have used AI as a fancy assistant — a tool that summarizes literature, screens molecules, or drafts code while a human directs the work. The Molecule.one campaign blurred that line. The model didn’t just help; it chose the problem (improving a reaction on a chemical building block found in more than 90 approved medicines), generated and ranked its own proposals, and ran the lab work largely on its own before human chemists confirmed the result at scale.
That matters far beyond pharma. Any business that runs research and development — biotech, medical devices, peptides and specialty chemicals, materials, even consumer-product formulation — is now staring at the same scenario. When the most valuable idea in your pipeline came primarily from a machine, your ability to protect it with a patent is no longer automatic. And a patent you can’t defend is an asset that quietly evaporates in financing, licensing, and especially in a sale.
In plain terms: the science got easier, and the ownership got harder. Below is what every business owner relying on AI in the lab — or in the codebase — needs to understand before the next breakthrough lands.
The Legal Impact: 5 Patent Risks of AI-Generated Inventions
U.S. law settled the headline question first. In Thaler v. Vidal, the Federal Circuit held that a patent inventor must be a natural person — an AI system cannot be named as an inventor — and the Supreme Court has now declined to revisit that conclusion in both the patent and copyright contexts. On March 2, 2026, the Court denied certiorari in Thaler v. Perlmutter, ending Dr. Thaler’s parallel bid to register AI-only artwork. The message from the courts is consistent: if you want intellectual property protection, a human has to be in the loop. Here is how that principle creates concrete risk for AI-generated inventions.
1. Your invention may have no valid inventor at all
On November 26, 2025, the USPTO issued its Revised Inventorship Guidance for AI-Assisted Inventions, which rescinded the agency’s 2024 framework in full. The new rule treats AI as a tool — like lab equipment or software — and asks one traditional question: did a natural person conceive the invention? Conception requires “a specific, settled idea, a particular solution to the problem at hand.” Critically, the USPTO says that having “a general goal or research plan” is not enough. If your scientists merely pointed the AI at a problem and the machine supplied the actual solution, you may not have a qualifying human inventor — and an application with no proper inventor is invalid.
2. Patents already filed could be vulnerable to challenge
Inventorship is not a formality. A patent obtained without naming the correct inventor — or by naming a human who didn’t actually conceive the claimed solution — can be attacked as invalid or unenforceable in litigation. As AI moves deeper into R&D, expect challengers and accused infringers to start asking, in discovery, exactly what the human contributed versus what the model produced. Companies that didn’t document the human conception story may find their strongest patents are their most fragile.
3. Trade-secret strategy suddenly matters more
If a breakthrough can’t clear the inventorship bar, patents aren’t the only path. Trade-secret protection doesn’t require a named human inventor — but it requires genuine secrecy: access controls, confidentiality agreements, and disciplined handling of AI prompts, outputs, and lab data. For many AI-generated inventions, the right answer will be a deliberate choice between patenting and protecting as a trade secret, made early, not after the work is published.
4. Your contracts probably don’t address AI outputs
Most invention-assignment clauses, vendor agreements, and joint-development deals were written for a world where humans invent and machines compute. They rarely say who owns what an AI tool generates, whether the AI vendor retains any rights, or how the human-contribution record will be kept. Employment agreements, independent-contractor terms, and AI software licenses all need fresh review so that ownership of AI-generated inventions is clear before a dispute — not litigated after one.
5. M&A and financing diligence just got a new line item
When a buyer or investor values a company, the patent portfolio is often the crown jewel. If those patents flowed from AI-assisted research without a documented human-conception trail, a careful acquirer will discount them — or demand indemnities. Expect AI-inventorship representations to appear in purchase agreements, and expect M&A diligence to probe how each key invention was actually made. This is also where a true inventorship fight can spill into court; contested-ownership and inventorship disputes are litigation matters handled by our colleagues at Howard Law Group.
What Howard East Clients Should Do Now
You don’t need to abandon AI in the lab — you need a paper trail and a plan. Three moves are worth making this quarter.
First, document human conception in real time. For any AI-assisted project, keep contemporaneous records of what your people actually decided: the specific solution they recognized, refined, or selected, not just the goal they set. Under the USPTO’s new standard, that record is the difference between a defensible patent and a void one.
Second, decide patent-versus-trade-secret early. Before publishing or filing, get counsel to assess whether a given AI-generated invention can meet the human-conception bar — and if it can’t, build the secrecy controls that trade-secret protection requires. The choice is hard to reverse once the work is public.
Third, update your contracts and policies. Refresh invention-assignment, vendor, and AI-tool agreements so ownership and record-keeping obligations are explicit. Businesses in regulated and specialized fields — from healthcare and life sciences to cannabis operators patenting extraction and formulation methods — face the inventorship question the moment AI touches their R&D, and they should treat corporate and IP hygiene as part of the cost of using these tools. For deeper background on how fast the ground is shifting, see our recent analysis of AI export controls and employer risk.
Frequently Asked Questions
Can an AI be listed as an inventor on a U.S. patent?
No. Under Thaler v. Vidal and the USPTO’s 2025 guidance, only a natural person can be a named inventor. AI is treated as a tool used by human inventors.
Are AI-generated inventions automatically unpatentable?
Not automatically. AI-assisted inventions can be patented if at least one natural person made a significant contribution to conceiving the specific claimed solution. The risk arises when no human conceived more than a general goal.
What should we do if our breakthrough came mostly from AI?
Talk to counsel about whether the human-conception standard is met. If it isn’t, trade-secret protection — backed by real secrecy controls — is often the better strategy. The right path depends on your facts.
This article is for informational purposes only and does not constitute legal advice, and reading it does not create an attorney-client relationship. AI-generated invention and inventorship questions are fact-specific and evolving; consult qualified counsel about your situation. Attorney Advertising.
Talk to Howard East About Protecting AI-Generated Inventions
If your business uses AI in research, product development, or engineering, now is the time to make sure your most valuable ideas are actually protectable. Howard East advises business owners on intellectual property strategy, contracts, and M&A readiness in the age of AI. Book a consultation to review your patent and trade-secret posture before your next breakthrough.
Source: Dr. Alex Wissner-Gross, The Innermost Loop, June 17, 2026.
