What changed

Anthropic's Frontier Red Team has shortened the window it gives software makers before publishing the security flaws it finds. For high- and critical-severity flaws, the waiting period drops from 90 days to 30. Other flaws get 45 days.

The rationale: AI finds bugs faster

According to the company, AI models now detect software bugs far more quickly than before. That makes it necessary to shorten the fix window offered to software vendors at a similar pace.

What it means for the industry

The disclosure timeline directly affects whether patches can keep up. A shorter window pushes vendors to ship fixes faster, while raising questions about how long users may remain exposed.