Over the past year, we’ve published lots of data that shows that frontier AI models are more than capable of discovering and exploiting security vulnerabilities. Showing how previously unknown ...