Anthropic accidentally leaked details of a new AI model that poses unprecedented cybersecurity risks

Anthropic has accidentally leaked internal documentation and early performance benchmarks for its next-generation AI model, 'Mythos.' The leak, which occurred via an unsecured staging server, reveals that Mythos possesses unprecedented capabilities in autonomous software exploitation and social engineering. Cybersecurity experts who reviewed the leaked data suggest the model can identify complex vulnerabilities in legacy infrastructure that were previously thought to be secure. The disclosure has prompted urgent calls for a security audit of the model's weights and a reassessment of AI safety protocols across the industry. Anthropic has since secured the data, but the information is already circulating in closed developer circles.