Multimodal AI Drafts Patents with 95% Accuracy — Legal Industry Faces Disruption
An AI system just generated 1,000 patent applications at near-human quality. IP law firms are scrambling to redefine what their associates actually do.
🔍 What Happened
A collaboration between Stanford and a leading IP law firm tested GPT-5 and Claude Opus on drafting 1,000 real patent applications. Expert reviewers rated 95% of outputs as 'acceptable without significant revision' and 67% as 'superior to typical associate work.' Results were published in the Journal of IP Law this week.
💡 Why It Matters
Patent drafting is a $4B/year industry dominated by law firms charging $400-800/hour. If AI can handle most drafting, the billing model collapses. Law firms must pivot from hourly work to strategic advisory and litigation — or lose market share to legal-tech startups.
🏢 Impact on Business & Users
Expect rapid consolidation in IP law. Big firms like Morrison Foerster and Fish & Richardson are building internal AI stacks. Meanwhile, legal-tech upstarts like Harvey and EvenUp are offering 10x cheaper drafting services. Corporate legal departments can now insource much more IP work.
👀 What to Watch Next
Watch for bar association responses — will unauthorized practice of law rules be enforced against AI-first platforms? Also track USPTO policy on AI-authored inventions. Some patent offices are already requiring human inventor disclosures.
Frequently Asked Questions
Enjoyed this article?
Get stories like this delivered to your inbox.
