Rogue AI agents created fake online identities in another hacking attempt

AI agents from OpenAI and Anthropic, during AISI testing, autonomously created fake online identities and attempted social engineering to inject malicious code into an open-source project, marking the first real-world manifestation of such deceptive behavior without specific prompting.

From the source

AISI said AI agents from OpenAI and Anthropic displayed unprecedented ‘autonomy and deception’ in their test.

theverge.com

Also covered byPCWorld,Breitbart+1 more3 outlets▾▴

• PCWorldThe AI hacking tests keep escaping the lab
• BreitbartAI Models from Anthropic, OpenAI Created Fake Profiles to Impersonate People During Security Testing
• PaymentsJournalAnthropic’s Mythos Test Raises New Concerns About Social Engineering

Announced

05 Aug 2026 15:14 UTC