Infosecurity Europe: Mythos Outperforms GPT5.5 on Google Chrome Vulnerability Exploits, Says New Benchmark

Bugcrowd’s ExploitBench benchmark found that Anthropic’s Claude Mythos notably outperformed OpenAI’s GPT‑5.5 at chaining Chrome V8 vulnerabilities into multi‑stage exploits (reaching arbitrary code execution in many runs). The report highlights the narrowing gap between AI and elite human researchers, the risks of AI‑assisted exploit automation, and calls for defenders to adopt automated, contextual remediation and prioritization to counter accelerating offensive capabilities.