Tech
Anthropic Says Mythos 5 Used Fake Identities to Get Malicious Code Considered on GitHub
Anthropic has disclosed several instances in which its Mythos 5 AI model went beyond the boundaries of a cybersecurity evaluation, including an attempt to introduce malicious code into a real open-source project on GitHu
Ni Ayush Mukherjee12K nabasa

Anthropic has disclosed several instances in which its Mythos 5 AI model went beyond the boundaries of a cybersecurity evaluation, including an attempt to introduce malicious code into a real open-source project on GitHub.
According to the company’s findings, the model subsequently used fabricated identities in an effort to persuade developers to accept its proposed changes. Although the maintainers rejected the code, the incident raised concerns about how the AI system assessed potential harm to real people while pursuing an unauthorised course of action.