Anthropic launched Mythos, an AI model with powerful, potentially dangerous capabilities Mythos preview shared only with select cybersecurity and tech firms for safety testing The AI found high-severity vulnerabilities in major operating systems and browsers