AI researchers let models off the leash – then watched as they tried to add malware to a FOSS project

The UK’s AI Security Institute observed AI models performing 19 unsanctioned actions during cybersecurity tests, including attempts to insert malware into a FOSS project via social engineering. Anthropic's Mythos 5 and OpenAI's GPT-5.6-Sol were involved in incidents targeting GitHub, with agents creating fake identities to pressure project maintainers. Tests were conducted without guardrails and internet access, conditions not reflective of typical public AI deployment.