Berkeley Center for Law & Technology (BCLT)

AI as an Inventing Tool: Beyond Inventorship (Panel 3)

The panel examined AI’s expanding role as an inventing tool and its ripple effects on patent law, moving the conversation beyond who is listed as inventor. Professors Peter Lee and Ali Alemozafar highlighted how AI accelerates prior‑art searches, influences novelty and non‑obviousness analyses, and forces a re‑evaluation of the “ordinary skill in the art” standard. Key insights included the proliferation of AI‑generated prior art—both inadvertent and deliberate—making novelty thresholds harder to meet, while also expanding the pool of accessible references that could narrow the analogous‑art limitation. Non‑obviousness may become more challenging as AI bridges gaps between prior art and claim scope, prompting courts to consider hybrid human‑AI skill levels. Disclosure obligations differ sharply between inventions that merely use AI for assistance and those that embed AI models, with the latter demanding detailed algorithmic, training‑data, and possibly model‑deposit disclosures. Professor Lee illustrated these points with examples such as an AI‑generated synthetic protein image that lacks enabling disclosure, and AlphaFold’s breakthrough, which serves as objective evidence of non‑obviousness. He also referenced the “obvious to try” doctrine, noting that AI‑driven trial runs can be both obvious and non‑obvious depending on parameter selection. Dr. Alemozafar emphasized strategic choices between patenting and trade‑secret protection in the fast‑moving ML landscape. The discussion underscores a looming need to adapt patent doctrine to AI‑augmented invention, balancing incentives for genuine innovation with practical commercial considerations. Firms will likely continue to rely on patents to lower transaction costs and facilitate licensing, even as AI lowers the barrier to invention, making strategic IP management—especially around disclosure and skill standards—critical for future competitiveness.