Anthropic opens Fable 5 to everyone: first model in the Mythos line goes public

Anthropic has released Fable 5, the first model in the new Mythos line available to the general public. Until today, Claude's most powerful model had been restricted to a small number of partners for two months because of cybersecurity concerns.

About the Mythos line

Mythos is a new generation of Claude models, introduced in April of this year. It is a significant leap forward in architecture, reasoning ability and reliability. Fable 5, as the line's flagship, delivered the best results on benchmarks for complex coding, multi-step logical analysis and long-context work.

Until today, Fable 5 was available only to a select group: universities, large corporations and research institutions. Anthropic used this period for intensive testing in real-world conditions and for improving its safety mechanisms based on the experience gained. The decision to open the model came after months of reviews and consultations with experts in AI safety and ethics.

Why it was closed to the public

Anthropic did not hide the reason: the most powerful models carry new risks that must be understood and managed before release into the world. The company analyzed potential misuse and identified several areas requiring heightened attention.

• Risks of social engineering and manipulation of users
• The possibility of generating malicious code and exploits
• Use in phishing and spam campaigns
• Automated creation of disinformation at scale
• Violations of confidentiality and privacy in sensitive areas
• Use in developing weapons and high-risk biotechnology

Two months of restricted access let the company gather data on how the model behaves in real-world use and adjust its protection algorithms.

Restrictions at launch

Fable 5 will be open to everyone, but with a clear set of restrictions. Anthropic will explicitly prohibit its use in developing nuclear, biological and chemical weapons, in synthesizing malware and in organizing large-scale information operations.

Beyond these absolute bans, the model will apply additional checks in higher-risk areas: cybersecurity, financial operations, the legal system and healthcare. When handling such topics, the model may ask for clarification of the context and purpose of a request.

"Safety is not the end of development; it is an integral part of it, built in from the start,"

Anthropic emphasizes.

What this means

Fable 5's public release demonstrates a model for the responsible release of powerful AI systems. Anthropic has shown that safety and accessibility do not necessarily conflict: it is possible to build cutting-edge tools while managing their risks at the same time. This is not the first such release, but it is an important precedent.

For developers, it means access to the most advanced Claude model. For companies, it brings new possibilities in automation, data analysis and development. For society, it shows that AI can be released deliberately, with potential consequences taken into account and protection mechanisms built in.