Key facts
- Anthropic is expanding access to its AI model, Mythos 5.
- Mythos 5 has no limitations on cybersecurity or biology.
- The model has helped users find over 10,000 critical security flaws.
- Access is currently limited to a select group of cyberdefenders and infrastructure providers.
- Anthropic plans to broaden access through a trusted program.
Anthropic is broadening access to its advanced AI model, Mythos 5, which has demonstrated significant capabilities in identifying security vulnerabilities. Initially provided to about 150 groups for preview, the model, which lacks limitations on cybersecurity or biology, has already been instrumental in discovering over 10,000 critical security flaws.
While current access is restricted to a small cohort of cyberdefenders and infrastructure providers, Anthropic stated its intention to expand availability through a "broader trusted access program." The company noted that Mythos, along with its similarly capable counterpart Fable, can operate "unattended" for extended durations, surpassing the capabilities of previous Claude models.
This development comes as Anthropic's private valuation approaches $1 trillion, underscoring investor confidence in its AI advancements. The move also occurs amidst ongoing discussions about AI safety, with Anthropic co-founder Jack Clark recently expressing concerns about the rapid pace of AI development and the need for mechanisms to control its advancement.