UK Government Sounds Alarm on Advanced Cyberattack Capabilities of GPT-5.5, Marks Second Threat Following 'Mythos'

On April 30, the AI Safety Institute (AISI), under the UK's Department of Science, Technology & Innovation, published an evaluation report indicating that OpenAI's latest model, 'GPT-5.5', possesses significant cyberattack capabilities. AISI assessed the cyberattack capabilities of GPT-5.5 using two different approaches. The first approach involved 'Capture The Flag' (CTF)...