A study by the Stanford Center for Legal Informatics shows that advanced AI reasoning models now match experienced patent attorneys in auditing patent claims. The systems evaluated draft patents against database records to find prior-art conflicts that could invalidate a application.
The benchmark tested 500 complex electrical and mechanical patent applications. The AI identified invalidating prior art with 91.2% accuracy, compared to 90.8% for a control group of human attorneys. However, the AI completed its audits in minutes, rather than the average 15 hours required by humans.
Redefining Legal Support Tasks
The results suggest a major shift in how law firms handle patent filings. Junior associates typically spend their first years doing research and review. With automation handling the search, firms can focus resources on writing strategy.
"This is not about replacing lawyers," says study co-author Sarah Jenkins. "It is about speed. An attorney can now review five times as many applications per month. It lowers the barrier for inventors looking to file patents."
Law firms are already integrating these models into their workflows to prepare for high-volume filing seasons.
