Anthropic Launches Claude Fable 5, Opens First Mythos-Class AI Model to Public
Advertisement
Alongside the model launch, Anthropic highlighted progress under Project Glasswing, a cybersecurity-focused initiative involving major technology and financial organizations.
Anthropic has launched Claude Fable 5, its most advanced artificial intelligence model to date and the first publicly available model in its new Mythos-class family. The announcement marks the broader release of technology that was previously available only through limited testing programs.
The launch comes two months after Anthropic introduced Claude Mythos Preview, an experimental version that demonstrated an unexpected ability to identify vulnerabilities across major operating systems and web browsers despite not being specifically designed for cybersecurity applications.
To address concerns about the potential misuse of highly capable AI systems, Anthropic has implemented safeguards in Claude Fable 5 that restrict responses in high-risk domains, including cybersecurity and biology. According to the company, users seeking information in those areas will instead be routed to its Claude Opus 4.8 model.
Alongside the model launch, Anthropic highlighted progress under Project Glasswing, a cybersecurity-focused initiative involving major technology and financial organizations. Participants include Amazon Web Services, Apple, Google, Cisco, Microsoft, JPMorgan Chase, and other partners.
As part of the program, Anthropic provided early access to Mythos Preview, allowing participating organizations to identify and remediate vulnerabilities within their software and systems before broader deployment.
The company said it conducted extensive internal and external testing to evaluate whether users could bypass the model’s safety controls. Anthropic also ran a bug bounty program involving more than 1,000 hours of testing and reported that researchers were unable to identify a universal jailbreak capable of consistently circumventing the model’s safeguards.
External organizations were also invited to test the model’s defenses, with Anthropic stating that those efforts similarly failed to compromise its core safety mechanisms.
In addition to launching Claude Fable 5, Anthropic said organizations that previously received access to Mythos Preview through Project Glasswing will be upgraded to Mythos 5, the latest version of the company’s restricted Mythos-class system.
Anthropic has priced access to both Claude Fable 5 and Mythos 5 at $10 per million input tokens and $50 per million output tokens.
The launch comes a week after Anthropic disclosed that it had confidentially filed paperwork for an initial public offering, signaling the company’s next phase of growth amid increasing competition in the artificial intelligence market.
Stay tuned for more such updates on Digital Health News
