CybersecurityAI Research11 min reading time

Path to Astra: critical capabilities and frontier safeguards

Open AI News
Read full post
Anthropic has assessed its AI model Astra as reaching a critical cybersecurity capability, able to find and exploit unknown security flaws autonomously. They have strengthened safeguards and will limit access to advanced features initially to testers before broader defensive use.

More in Cybersecurity

Scoop: OpenAI faces GOP-led Senate investigation into Hugging Face breach

Covered by 2 sources
Cybersecurity6 min read

Anthropic Discloses Fourth AI Hacking Incident Involving Claude Opus 4.6

Covered by 2 sources
Cybersecurity5 min read

Every AI Incident Has Two Timelines. We Default To One

Forbes