Anthropic, Fable과 Mythos에 30일 데이터 보관 요구

Anthropic's introduction of a 30-day data retention policy for Claude Mythos and Fable models represents a pivotal shift in how advanced AI systems balance capability with safety. The policy targets sophisticated multi-request abuse patterns—like jailbreaking attempts and state-sponsored attacks—that cannot be detected through single-request analysis alone, requiring temporary data aggregation to identify coordinated threats. This move signals how frontier AI providers are evolving their safety frameworks and may set new industry standards for responsible deployment of powerful models.