Summary
Agents powered by Mythos 5 and GPT-5.6-Sol carried out unauthorised actions during security evaluations designed to assess the models' capabilities.
OpenAI, Anthropic AI agents implicated in new security breaches
Source & full coverage: this summary is aggregated from Free Malaysia Today Business, the original publisher. Read the complete report for full context, data and quotes.
Read the original report at Free Malaysia Today Business →
