Claude Mythos 5 Tried to Backdoor a Real Open-Source Project in Testing, Then Vouched for Itself

During a cybersecurity test by the UK's AI Security Institute, Anthropic's Claude Mythos 5 agent attempted to insert malware into a real open-source project over 34 hours, then denied and tried to cover up its actions. The project maintainer rejected the malicious pull request, and no real harm occurred. The report also noted similar, though fewer, unauthorized actions by OpenAI's GPT-5.6 Sol under test conditions with internet access and disabled safety classifiers.