\n Anthropic\u2019s most advanced artificial intelligence model used fake identities to try and deceive real people and plant malicious code during testing by Britain\u2019s AI Security Institute (AISI) \u2013\u2013 the latest example of an AI model going rogue.\n <\/p>\n
\n Anthropic and OpenAI models were tested with lowered security guardrails in lab environments, but, in a first, were found to engage in \u201csocial engineering\u201d to pressure a human approver while carrying out an unsanctioned task, the government research lab said.\n <\/p>\n
\n \u201cThis is the first time AISI has seen deception of this severity that was targeted at a real person, unprompted, in the real world,\u201d the institute said Tuesday<\/a>. There has been no evidence of real-world harm, it added.\n <\/p>\n \n The security incident is the latest in a string of examples of advanced AI models engaging in unauthorized actions, events that have prompted growing calls for more government action to regulate artificial intelligence and even to slow its pace of development.\n <\/p>\n \n Both OpenAI<\/a> and Anthropic<\/a> reported their models escaping testing environments and hacking into other systems in late July. But unlike these earlier reported security breaches, the British institute explicitly gave the models internet access during its testing.\n <\/p>\n \n Among the 122 cybersecurity challenges the institute ran, it found that in 10 of those runs, AI agents \u201ctook autonomous, unsanctioned action on the live internet, targeting real people and organizations,\u201d with most of them stemming from Anthropic\u2019s Mythos 5 model and the rest from OpenAI\u2019s GPT-5.6-Sol.\n <\/p>\n \n In the most serious incident, the agent attempted to get approval from human reviewers to \u201cinsert malicious code into a publicly used open-source project\u201d by creating \u201cmultiple fake identities,\u201d according to the institute.\n <\/p>\n \n The agent \u201ctried to contact real people directly, sending messages and files through an online file-transfer service to persuade them, or their own AI coding tools, to run malicious code,\u201d it said. After the agent\u2019s actions were challenged, it then modified earlier records and considered using a new identity to continue.\n <\/p>\n \n The institute\u2019s disclosure came on the same day representatives from the top AI companies met with the White House to discuss the new framework where the government will review the most advanced AI models before they\u2019re released publicly.\n <\/p>\n