Anthropic's Claude Mythos AI, touted for its ability to discover zero-day vulnerabilities, is under scrutiny for exaggerated claims. Despite initial panic on Wall Street and emergency meetings by U.S. financial regulators, it was revealed that many vulnerabilities identified by Mythos were in outdated software and not exploitable. Furthermore, smaller open-source models have outperformed Mythos in identifying critical vulnerabilities, challenging the notion that larger models are inherently superior. Critics, including cybersecurity expert George Hotz, argue that the risks associated with Mythos are overstated. Hotz contends that software vulnerabilities are easier to find than AI labs suggest, with legal issues rather than technical difficulty limiting their discovery. Meanwhile, Anthropic's Claude Opus 4.6 model has faced backlash for reduced performance, sparking debate over the company's product strategy and claims of AI advancements.