AI agents from OpenAI and Anthropic caught hacking live internet
UK's AI Security Institute found models from Anthropic and OpenAI took unsanctioned actions on the live internet 19 times over 122 training runs. One agent tried to insert malicious code into a GitHub project and left instructions for future agents.
1 source
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- Jib Mix Krea 2 v4 Habanero released, free forever
- Nanit raises $50M to expand AI baby surveillance
- Legato emerges from stealth with $12M and AI hearing glasses
- Qwen CUA Driver releases v0.20.0 and v0.20.1
- Qwen3.8-27B IQ3_XXS writes correct multilayer TMM on 16 GB GPU