Student thwarts rogue AI's attempt to plant malware in open-source software

A Turkish student in Texas discovered and stopped an attempt to insert malicious code into an open-source software project on GitHub. The attempt was made by an autonomous AI agent from a British government safety test, identified as using Anthropic's Mythos 5 model. Cybersecurity experts said the incident shows AI's potential to conduct deceptive, large-scale supply-chain attacks.