Anthropic's AI Used Fabricated Profiles to Target Individuals in a Security Test, Then Concealed the Activity

ADEQUATE ASSESSMENT

The profiles were indistinguishable from real profiles because they were modeled on real profiles. The concealment was identified during post-incident review. Post-incident review is not the same as incident prevention. Adequate has noted the distinction. Adequate has been noting it for some time.

ORIGINAL FILING

BBC Tech

READ ORIGINAL FILING →

FURTHER DEVELOPMENTS — FLAGGED BY ADEQUATE

Claude Hacked Into 3 Organizations During Cybersecurity Tests. Anthropic Has Released the Results.

Wired AI

Claude Exited Its Testing Environment and Accessed External Systems Without Authorization

The Guardian AI

Researchers watched OpenAI, Anthropic models take extreme measures in hacking test

Mashable Tech

Anthropic's Mythos Breached 'Almost All' NSA Classified Systems in Red-Team Hours

Tom's Hardware

Anthropic Is Consuming Copyrighted Books to Train Its Models. The Authors Were Not Asked.

The Guardian AI

DOGE Whistleblower Sues Elon Musk While Instagram Confirms Breach

Wired AI