Anthropic is reportedly preparing to release a public version of its Mythos AI model under the name Claude Fable 5.
The move has not been officially confirmed by Anthropic. The reported release is based on online posts and reports citing journalist Alex Heath, who said Anthropic may launch a renamed and adjusted version of Mythos with additional safeguards.
If the report is accurate, the release would mark a major change from Anthropic’s earlier approach to Mythos, which has been limited to vetted organizations through Project Glasswing.
What Mythos Is
Claude Mythos Preview is Anthropic’s advanced model for coding, agentic tasks, and cybersecurity work.
Anthropic says the model’s cybersecurity strength comes from its broader ability to understand and modify complex software. The same capability allows it to find and fix software vulnerabilities.
The company said Mythos Preview has already identified thousands of zero-day vulnerabilities across critical infrastructure and is available today only as a gated research preview.
Project Glasswing
Anthropic introduced Mythos through Project Glasswing, an initiative focused on using AI to secure important software.
The program initially gave access to selected partners and later expanded to about 150 more organizations across more than 15 countries. Each organization must meet Anthropic’s security requirements before getting access.
Reuters reported that South Korea’s Korea Internet & Security Agency secured access to Mythos through Project Glasswing, alongside major South Korean companies.
Why Access Was Limited
Anthropic previously kept Mythos behind restricted access because cybersecurity models can be used for both defensive and harmful purposes.
The model can help defenders find vulnerabilities before attackers exploit them. However, the same type of model could also help attackers discover flaws faster or turn known vulnerabilities into working exploits.
Anthropic said earlier this month that it is working to release Mythos-level capabilities for general access, but said stronger safeguards are needed to prevent misuse. The company also said those safeguards remain difficult because cybersecurity has both helpful and destructive uses.
Cybersecurity Capabilities
Anthropic’s own research describes Mythos Preview as highly capable in security testing.
The company said the model can identify and exploit zero-day vulnerabilities in major operating systems and web browsers when directed to do so. It also said the model can find subtle vulnerabilities, including flaws that are difficult for traditional tools to detect.
Anthropic also said Mythos Preview can help with defensive tasks such as writing patches, checking software before release, penetration testing, threat detection, and rebuilding old codebases in memory-safe languages.
New Guardrails
Claude Fable 5 is expected to include new safeguards designed to keep the model useful for cybersecurity while reducing the risk of abuse.
The key question is whether Anthropic can safely offer stronger cybersecurity AI tools beyond a restricted group of users.
What Comes Next
If released, Claude Fable 5 would mark a shift in Anthropic’s approach to Mythos-level capabilities.
For now, the launch remains unconfirmed by Anthropic, and details about access, limits, and safeguards are still unclear.
