src=”https://techcrunch.com/wp-content/uploads/2026/06/GettyImages-2279025960.jpg” />
Anthropic Releases Advanced AI Model for Public Access with Enhanced Safety Protocols
Anthropic has made its most powerful AI model, Claude Fable 5, available to the general public for the first time, integrating significant safety measures to mitigate potential risks. This release marks a critical step in making cutting-edge AI capabilities accessible while addressing concerns surrounding its deployment.
Core Technology and Safety Framework
Claude Fable 5 is the initial public iteration of Anthropic’s advanced Mythos model. The company reports that Fable 5 demonstrates exceptional proficiency in software engineering, complex knowledge work, and visual processing. However, to ensure responsible use, the model incorporates stringent safety limitations. In sensitive domains such as cybersecurity, biology, chemistry, and distillation, Fable 5 is programmed to prevent responses and revert to the capabilities of Claude Opus 4.8, a more established model.
Phased Rollout and Enterprise Access
Initially previewed in April and subsequently limited to a select group of partners due to security considerations, the Mythos model saw expanded access last week to hundreds of organizations across 15 countries, with a particular focus on entities managing critical infrastructure. Now, a version of this advanced technology is accessible via Anthropic’s Claude API and its consumption-based Enterprise plans. The integration into subscription services is being rolled out incrementally: through June 22, Fable 5 will be included in Pro, Max, Team, and seat-based Enterprise plans without additional cost. Following this period, from June 23, Fable 5 will transition to a usage-credit system, with plans to reinstate it as a standard subscription feature in the near future.
Concurrently, Anthropic is deploying a new iteration, Mythos 5, to organizations that have already received approval for access to the advanced model.
Strategic Positioning and Safety Concerns
The launch of Fable 5 occurs as Anthropic prepares for potential public market entry, a move that mirrors developments with other major AI players. This release also follows Anthropic’s recent call for leading AI laboratories to collaborate on establishing collective safety mechanisms for frontier AI development. The company has voiced concerns that the rapid advancement of AI systems could lead to recursive self-improvement (RSI), where models autonomously enhance their capabilities without human oversight.
Acknowledging the potential risks associated with a powerful model like Mythos falling into the wrong hands, Anthropic conducted extensive stress-testing of its safety classifiers against simulated “jailbreak” attempts prior to Fable 5’s release. The company reported that an internal bug bounty program and engagements with external red-teaming organizations failed to yield universal jailbreaks after more than 1,000 hours of testing.
Despite these measures, the potential for novel attack vectors remains. Consequently, with the introduction of both Fable 5 and Mythos 5, Anthropic is instituting a mandatory 30-day data retention policy for all traffic, even for enterprises that previously held zero-retention agreements. Anthropic asserts that this data will not be used for training purposes but exclusively for defending against sophisticated and emerging threats, including new jailbreak techniques, and for refining the accuracy of its safety classifiers. This policy could establish a new industry standard where access to highly advanced AI models is contingent upon adherence to data retention protocols framed as essential safety measures.
Performance and Market Reception
For users engaging with the model, not every query will necessitate the full capabilities of Fable 5. Anthropic indicates that instances requiring the model to defer to Opus 4.8 are infrequent, with initial data suggesting that over 95% of Fable sessions operate entirely on its native responses. Third-party evaluations have highlighted Fable’s performance: analytics firm Hex reported that Fable was the first model to achieve a 90% score on its benchmark for complex, long-running analytical tasks, noting its “strong judgment and attention to nuance” on challenging questions. The Vibe-coding platform Base44 observed Fable’s superior capability in “one-shotting full apps” and its advanced tool-calling features. Genspark, an AI-powered workspace and agent platform, found Fable to outperform all other models in its evaluations, particularly in tasks related to UI design and game coding.
Pricing and Economic Considerations
The pricing structure for both Fable 5 and Mythos 5 is set at $10 per million input tokens and $50 per million output tokens, effectively doubling the cost of Claude Opus 4.8. This premium pricing may act as a natural constraint on widespread adoption, especially as many enterprises are increasingly scrutinizing AI expenditures following significant early-season budget depletion or unexpected cost overruns. Advanced models, with their sophisticated reasoning abilities that can break down complex requests into multiple sub-tasks, can amplify these cost concerns.
Anthropic anticipates substantial and potentially unpredictable demand for Fable 5. Some organizations, such as the shopping rewards platform Rakuten, perceive the potential benefits to outweigh the higher cost, stating that Fable’s ability to “reflect on and validate its own work” enables “highly autonomous operations,” where the enhanced analytical effort proves financially advantageous.
Business Style Takeaway: Anthropic’s release of Fable 5 underscores the ongoing tension between AI capability advancement and the imperative for safety and cost management. Businesses must now evaluate whether the performance gains of premium models justify the increased investment and potential data retention requirements, particularly as the industry grapples with establishing robust safety standards for increasingly powerful AI systems.
Learn more at : techcrunch.com