AI models surpass cybersecurity autonomy benchmarks, researchers say (opens in new tab)

Researchers reported a sharp jump in autonomous cyber capability, citing separate findings from the United Kingdom’s AI Security Institute and Palo Alto Networks that Anthropic’s Claude Mythos Preview and OpenAI’s GPT-5.5 outperformed the cyber-capability trend lines researchers had been tracking.Highlights:Exploit code: T3N reported that models such as Claude Mythos can not only find software vulnerabilities but also write code that exploits them.Perspectives:Anthropic: Anthropic framed Clau...