Anthropic (NASDAQ:ANTP), developer of the Claude family of artificial intelligence models, has launched a public version of its advanced Mythos-class technology while retaining restrictions on sensitive use cases.
The newly released model, called Fable 5, marks the first broadly available offering from Anthropic’s Mythos series, the company’s most advanced generation of AI systems. The Mythos platform was unveiled in April but had remained limited to a small group of approved institutions due to concerns surrounding cybersecurity and misuse.
Anthropic said Fable 5 is designed to support tasks such as software development, complex research, image analysis and technical problem-solving.
Restricted Version Remains Available to Selected Partners
Alongside the public release, Anthropic is continuing to provide an unrestricted version known as Claude Mythos 5 to a select group of organisations with authorised access to the platform.
These users include cybersecurity partners participating in the company’s Project Glasswing initiative. Anthropic expanded the programme earlier this month to roughly 200 organisations across more than 15 countries and expects participation to continue growing.
The company has previously argued that Mythos possesses the capability to rapidly identify vulnerabilities in critical infrastructure, including financial networks and power systems, prompting a more cautious rollout strategy.
Security Concerns Shaped the Launch Strategy
Anthropic’s decision to limit access to Mythos initially generated debate within the technology sector, with some critics questioning whether the company was overstating potential risks.
However, organisations that have tested the model have generally supported Anthropic’s assessment of its capabilities. The U.S. government has also conducted evaluations of the technology as part of broader security reviews.
Following those assessments, the White House established a framework under which leading AI developers can collaborate with government agencies to evaluate their most advanced models prior to public deployment. Anthropic said Mythos 5 is being deployed under that arrangement.
Sensitive Queries Redirected to Less Powerful Models
To reduce potential risks, Anthropic has implemented safeguards within Fable 5 that redirect many requests involving cybersecurity, biology or chemistry to Claude Opus 4.8, a less capable model released publicly in May.
The company said similar restrictions apply to suspected attempts to extract information that could be used to train competing AI systems, particularly where concerns exist around unauthorised technology transfer.
Anthropic also conducted extensive security testing before launch, hiring independent specialists to spend more than 1,000 hours attempting to bypass the model’s protections through a process known as red-teaming.
In addition, the company ran a bug bounty programme that rewarded researchers for identifying vulnerabilities. According to Anthropic, no participant succeeded in fully circumventing the safeguards.
AI Policy Disputes and Commercial Challenges Continue
The release comes amid ongoing tensions between Anthropic and the U.S. administration over the company’s stance on applications involving mass surveillance and autonomous weapons systems.
The disagreement led to the termination of Pentagon contracts previously held by Anthropic, despite its AI systems having secured defence-related security clearances.
Fable 5 is also being introduced at a premium price point. Anthropic said usage will cost $10 per million input tokens and $50 per million output tokens, approximately double the pricing of Claude Opus 4.8.
Despite strong revenue growth, the company remains unprofitable and continues to face significant infrastructure costs. Anthropic recently entered into an agreement to lease data centre capacity from Elon Musk’s xAI, part of SpaceX, at a reported cost of $1.25 billion per month.
AI Sector Momentum Remains Intense
The launch of Fable 5 arrives during a period of heightened investor interest in artificial intelligence companies.
Both Anthropic and OpenAI recently disclosed plans related to potential public offerings, while SpaceX is preparing for a highly anticipated market debut that could become one of the largest listings in history.
As competition intensifies across the AI sector, Anthropic is positioning its Mythos platform as a powerful but carefully controlled technology aimed at balancing innovation with security considerations.
More about Anthropic
Anthropic is a San Francisco-based artificial intelligence company focused on developing advanced AI systems with an emphasis on safety, reliability and responsible deployment. The company is best known for its Claude family of large language models and competes with major AI developers including OpenAI, Google and xAI in the race to build next-generation artificial intelligence technologies.