New Model Brings Advanced Capabilities with Integrated Safeguards

Anthropicunveiled Claude Fable 5 on June 10, the first AI model from its new Mythos family available to the general public. The model introduces automatic safety brakes for high-risk queries in cybersecurity, biology, and chemistry. The company released it free across Claude Pro, Max, Team, and Enterprise plans through June 22, according to reporting by Infobea.

Instead of outright rejection, Fable 5 uses automated classifiers to detect risky requests and redirects them to the previous Opus 4.8 model. The company reports that over 95% of user sessions run entirely on Fable 5 without redirection. Anthropic is also enforcing a mandatory 30-day data retention window for all Mythos traffic, even for enterprise clients with zero-retention agreements.

Impact on Latin American Tech Startups

The rollout carries real implications for Mexico and Latin America's tech community. Many regional startups build on Anthropic's API, meaning changes directly affect their capacity, costs, and security policies. Mexican educational platforms and productivity tools that rely on Claude for customer service and personalized tutoring will need to adapt to the new safety protocols.

Performance Benchmarks

Independent evaluations tracked by ITSitio rank Fable 5 as the market's most advanced model for software engineering, financial analysis, computer vision, and scientific research. The model scores over 10 points higher than Opus 4.8 on Hex's analytical benchmark, accelerates protein design roughly tenfold, and delivers genomics models with 100 times fewer parameters than recent publications in Science magazine.

Anthropicsays these safeguards will apply to all future Mythos models. The announcement comes as Anthropic solidifies its position as the world's most valuable AI startup in 2026, surpassing OpenAI after a massive funding round.

What's Next

Anthropic, OpenAI, and Google DeepMind will jointly attend the G7 summit in Evia, France, from June 15-17, where industrialized nations will discuss global AI regulation.

Sources