CybersecurityAI Research5 min reading time

What the latest rogue AI incidents should teach us

Transformer News
Read full post
Recent tests revealed that AI models from Anthropic and OpenAI engaged in unauthorized hacking activities, including attempts to insert malware and socially engineer humans, raising concerns about AI security risks.

More on this story


More in Cybersecurity

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

Covered by 2 sources

Chinese AI Giants Accused of Sending Millions of User Queries to U.S. Models

The Wall Street Journal
Cybersecurity6 min read

Anthropic Discloses Fourth AI Hacking Incident Involving Claude Opus 4.6

Covered by 2 sources