Anthropic's latest AI model, Claude Opus 4.6, found a Firefox bug for the first time within 20 minutes during internal testing, the Wall Street Journal (WSJ) reported on Thursday.
After the Anthropic team submitted the bug, Firefox developers responded immediately.
Mozilla said Claude found more high-risk Firefox bugs in a two-week test in January than are typically reported worldwide over two months, WSJ reported.
Anthropic found more than 100 bugs over two weeks, including 14 classified as high risk. If appropriate attack code had been developed, the bugs could have been abused for wide-ranging attacks targeting all Firefox users, WSJ reported. Firefox patched 73 high-risk and critical bugs over the past year.
The test began with an idea from Anthropic's security team to try deploying Claude on complex browser software.
Chi-gyu Hwang
delight@d-today.co.kr