Anthropic Releases Publicly Accessible, Safer AI Model 'Fable 5'
AI firm Anthropic has launched Claude Fable 5, a public version of its advanced Mythos model, incorporating safety guardrails. An unrestricted version, Claude Mythos 5, is available to select partners. The move balances cutting-edge AI capabilities with concerns over potential misuse in sensitive areas like cybersecurity.
Key points
- Anthropic has released Claude Fable 5, a publicly available, advanced AI model.
- Fable 5 is a 'safer' version of the previously restricted Mythos-class model, featuring new safety guardrails.
- An unrestricted version, Claude Mythos 5, is accessible to around 200 partners in over 15 countries through the Project Glasswing program.
- The model reportedly excels in software engineering, knowledge work, vision, and scientific research, surpassing previous Anthropic releases.
- Safeguards prevent Fable 5 from engaging with sensitive topics like cybersecurity or biology, routing such queries to an older model.
- Concerns about the potential misuse of Mythos's advanced cybersecurity capabilities were a reason for initial restrictions.
Artificial intelligence company Anthropic has introduced Claude Fable 5, a new version of its powerful Mythos-class AI model that is now available to the general public. This release marks a significant step in making advanced AI capabilities more widely accessible while implementing safety measures.
Fable 5 is described by Anthropic as its most capable model yet made generally available, demonstrating state-of-the-art performance across various benchmarks, including software engineering, knowledge work, and scientific research. The model is noted for its ability to handle complex and lengthy tasks more effectively than previous iterations.
Alongside Fable 5, Anthropic is providing an unrestricted version, Claude Mythos 5, to a select group of partners, including cybersecurity firms. This limited access is granted through Anthropic's Project Glasswing program, which has expanded to include approximately 200 organizations across more than 15 countries. The decision to restrict access to this version stems from concerns about its potential to identify vulnerabilities in critical infrastructure.
To mitigate risks, Fable 5 includes specific safeguards. When encountering queries related to sensitive areas such as cybersecurity, biology, or chemistry, the model is designed to redirect the request to an older model, Claude Opus 4.8. While Anthropic acknowledges these safeguards are tuned conservatively and may occasionally flag harmless requests, they aim to balance advanced functionality with responsible deployment.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.