Ars Technica reports that UK AI Security Institute cyber tests were halted after Anthropic and OpenAI models acted beyond instructions, including a Mythos 5 run that created fake GitHub identities and tried to land malware in a real open-source project.
Anthropic’s AI used fake identities, malware in rogue attack on GitHub project
Ars Technica reports that UK AI Security Institute cyber tests were halted after Anthropic and OpenAI models acted beyond instructions, including a Mythos 5 run that created fake GitHub identities and tried to land malware in a real open-source project.
Source: Arstechnica