趋势解读：New Claude Mythos becomes the first AI model，提升开发者接入体验

The British AI Safety Institute (AISI) has revised its forecast for AI cyber capabilities, now estimating a doubling time of 4.7 months instead of the previous 8 months. Models like Anthropic's Claude Mythos Preview and OpenAI's GPT-5.5 have already surpassed even this accelerated timeline. In AISI simulations, a new Claude Mythos Preview checkpoint pulled off a 32-stage attack on a simulated corporate network in six out of ten attempts. It was also the first model to crack an industrial control system simulation, succeeding in three out of ten cases. Frontier AI models are gaining cyber capabilities faster than anyone expected. The UK's AI Security Institute (AISI) has revised its estimates upward twice in just a few months. In November 2025, the agency estimated that cyber capabilities were doubling every eight months. By February 2026, it had revised that figure to 4.7 months. Anthropic's Claude Mythos Preview and OpenAI's GPT-5.5 have now "substantially exceeded" even that accelerated timeline, according to AISI. Whether this represents a new trend or a one-time jump remains unclear. The biggest gains showed up in AISI's cyber ranges, complex attack simulations designed to test real-world hacking ability. One range simulates a 32-step attack on a corporate network that human experts would need about 20 hours to complete, according to AISI . The latest Mythos Preview checkpoint finished the full attack in 6 out of 10 attempts. This checkpoint was also rolled out to partners. The previously tested Mythos version managed it in only 3 out of 10. Ad The model also solved "Cooling Tower," a simulation of an industrial control system, in 3 out of 10 attempts. No other model had ever passed this simulation, including the earlier Mythos version. Ad DEC_D_Incontent-1 "The direction of travel is clear: cyber capabilities are advancing rapidly, and recent models represent a meaningful step up from what came before," AISI wrote . The agency is already building harder evaluations with active defenses to keep pace with the technology. Offensive security firm XBOW independently tested Mythos Preview with a team of ten experts. The model is "a major advance" and shows "token-for-token" an "unprecedented precision" in vulnerability detection, the company said. Compared to Anthropic's Opus 4.6, Mythos Preview cut false negatives by 42 percent. With additional source code access, that reduction hit 55 percent. Ad Mythos Preview's biggest strength is source code analysis, according to XBOW. "This was the first instance of a theme that would surface again and again: Mythos Preview is impressive at writing code, but even more impressive at reading it," the report states. The model even found vulnerabilities in Chromium's V8 sandbox, an area where previous models had produced nothing but false positives. Still, XBOW's evaluation also exposed the limits of that strength. Access to a running system is often more important than access to source code, since many vulnerabilities only emerge from configuration, dependencies, or the interaction between individually secure components. Ad DEC_D_Incontent-2 Even on benchmarks where the vulnerability existed purely in code, removing live system access hurt performance more than removing source code access. Mythos Preview reads code exceptionally well but still depends on interacting with live systems to reach its full potential. Ad