Anthropic has released Claude Fable, the first model to carry the company's new Mythos-class designation. The release marks a notable shift in how Anthropic is structuring its model lineup, introducing a tier that sits above the existing Claude 3 and Claude 3.5 families. Fable arrives alongside a set of policy changes that have already drawn scrutiny from researchers and enterprise customers alike.
What Is the Mythos Class?
The Mythos designation appears to signal a new generation of models that Anthropic is positioning at the high end of capability. Anthropic had signaled the Mythos tier was coming in earlier communications, but details remained sparse until now. Claude Fable is the first concrete product to carry that label, and it represents the company's answer to escalating competition from OpenAI, Google, and Meta in the frontier model space. Coverage from The Verge confirmed the release, describing it as Anthropic's entry into a new model category the company has been developing internally for some time.
Key Facts
- Claude Fable is the first model released under Anthropic's Mythos-class tier.
- Data collection is mandatory for Fable users and cannot be opted out of.
- Anthropic walked back a policy that researchers said could have interfered with AI safety work.
- Enterprise zero-data-retention arrangements are affected by the new policy framework.
- The Mythos class is positioned above existing Claude 3 and Claude 3.5 model lines.
The launch has not been without friction. Two separate policy issues surfaced in close proximity to the release. First, WIRED reported that Anthropic had initially introduced a policy that AI researchers said could have undermined their ability to use Claude for safety-related work. Anthropic subsequently reversed that policy. The reversal suggests the company received significant pushback before the change went into full effect, though Anthropic has not detailed the exact scope of what was rolled back or why it was introduced in the first place.
The policy, as initially written, could have prevented researchers from using Claude in ways that are fundamental to independent AI safety evaluation.WIRED
Data Collection and the Enterprise Question
The second controversy involves data handling. Claude Fable collects user data by default, and there is no opt-out mechanism, according to reporting from Mashable. That stands in contrast to arrangements some enterprise customers have operated under, where zero-data-retention agreements gave organizations greater control over what Anthropic retains from their interactions. The shift effectively ends zero-data-retention as an option for Fable users, raising compliance questions for businesses operating in regulated industries such as healthcare, finance, and legal services.
For individual users, mandatory data collection is common practice across AI platforms. For enterprise clients who negotiated specific data terms, the change is more disruptive. Anthropic has not issued detailed public guidance on how existing enterprise contracts will be handled during the transition to Fable, and it is unclear whether legacy arrangements will be honored or phased out.
Where Fable Fits
Taken together, the Fable launch illustrates how quickly the conditions around a new model release can become complicated. The capability story is straightforward enough: a new tier, presumably more powerful, arriving to compete with frontier offerings from rival labs. The policy story is messier. Anthropic has had to walk back at least one significant restriction, and the data collection terms have drawn enough attention to warrant coverage from multiple outlets on the same day as the launch itself.
The Mythos-class release includes safety guardrails that Anthropic says are built into the model's design rather than layered on afterward. Whether those guardrails satisfy the research community that objected to the initial policy remains an open question. Anthropic's safety commitments have always been central to the company's identity, and this release tests how that identity holds up when commercial pressure and researcher needs pull in different directions.
For users evaluating whether to adopt Fable, the calculus now includes not just capability benchmarks but data terms and policy stability. Those factors matter more than they once did, particularly for organizations that treat AI access as part of a broader compliance infrastructure. How Anthropic navigates the remaining questions around enterprise data and researcher access will likely shape how the Mythos class is received beyond the initial launch window.
