Anthropic has successfully restored Claude Fable 5 and launched Mythos 5 after the Trump administration lifted export controls that had previously restricted these models' availability. The move represents a significant shift in regulatory treatment for Anthropic's advanced AI systems, which had faced scrutiny over safety concerns. The company has responded to earlier restrictions by implementing refined guardrails designed to mitigate specific security risks while maintaining model functionality for legitimate use cases.
The core safety mechanism centers on a streamlined filtering approach: a single consolidated filter now blocks prompts that could enable users to identify software vulnerabilities and write code to exploit them. This targeted approach replaces more extensive restrictions while maintaining Anthropic's commitment to responsible AI development. The implementation reflects the company's ongoing work in Constitutional AI and safety research, demonstrating how technical safeguards can balance capability with risk mitigation without wholesale model restrictions.
The restoration underscores Anthropic's position in navigating regulatory dynamics around advanced AI systems. By deploying more sophisticated safety mechanisms rather than simply removing features, the company signals confidence in its ability to operationalize Constitutional AI principles in production systems. This approach may establish a template for how AI developers can address governmental security concerns while preserving model utility—a critical consideration as regulatory frameworks around advanced AI systems continue evolving globally.
Separately, Anthropic has introduced Claude Sonnet 5 with updated pricing structures, continuing the company's practice of adjusting costs while maintaining consistent token rates. The company simultaneously launched Claude Science, an AI-driven pharmaceutical R&D tool, expanding Claude's application into specialized scientific domains and demonstrating broader enterprise adoption momentum beyond general-purpose AI assistance.