GPT-5.5 matches heavily hyped Mythos Preview in new cybersecurity tests
New cybersecurity tests by the UK's AI Security Institute show that OpenAI's GPT-5.5 performs similarly to Anthropic's restricted Mythos Preview model on advanced cyber tasks. Both models achieved comparable success rates on Expert-level Capture the Flag challenges and the TLO simulation, with GPT-5.5 solving a complex Rust disassembler task efficiently. The findings suggest that the cybersecurity capabilities demonstrated are likely due to general AI advancements rather than a unique breakthrough in any single model.
Opening excerpt (first ~120 words) tap to expand
testing the mythos mythos GPT-5.5 matches heavily hyped Mythos Preview in new cybersecurity tests New results suggest Mythos’ cyber threat isn’t “a breakthrough specific to one model.” Kyle Orland – May 1, 2026 11:32 am | 25 An actual photo of GPT-5.5 performing AISI's cybersecurity tests. Credit: Getty Images An actual photo of GPT-5.5 performing AISI's cybersecurity tests. Credit: Getty Images Text settings Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only Learn more Minimize to nav Last month, Anthropic made a big deal about the supposedly outsize cybersecurity threat represented by its Mythos Preview model, leading the company to restrict the initial release to “critical industry partners.” But new research from the UK’s AI Security…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Ars Technica - All content.
Discussion
0 commentsMore from Ars Technica - All content
