Anthropic’s decision to open its once-restricted Claude Fable 5 model to the public crystallizes a core tension in AI: can a company scale cutting‑edge capability while credibly claiming it has defused its own worst risks?

On one side, the company and its boosters frame the move as a managed “race to the top.” CNBC highlights Anthropic’s pitch that Fable 5 is a “Mythos-class model” now offered to enterprise customers and paid subscribers, enabled by “new safeguards that block responses in specific high-risk areas, including cybersecurity and biology.”¹ The company emphasizes that Fable 5 delivers “exceptional performance” on software engineering and knowledge work, even scoring more than 10% higher than its earlier Claude Opus 4.8 on some benchmarks, while using guardrails that fall back to that weaker model when users probe for harmful content such as instructions to make toxins.¹

A parallel liberal perspective, drawing on reporting in the Guardian, stresses the safety narrative but casts it in more geopolitical and regulatory terms. Fable 5 is described as the “first to be made widely available from the company’s new Mythos class,” after months of restricted access over cybersecurity concerns, with Anthropic explicitly limiting its use in areas like critical infrastructure analysis.² At the same time, a more powerful, less-restricted Claude Mythos 5 is being deployed with select partners – including the US government – via Project Glasswing, part of a broader White House arrangement to test cutting-edge models before public release.²

Where both perspectives converge is in recognizing that safety is also strategy. CNBC underscores that Anthropic is “capitalizing on growing momentum and investor interest ... ahead of a potentially massive IPO,” linking guardrails to market positioning.¹ The Guardian notes Anthropic’s warnings about attempts from “authoritarian countries” to extract its technology, another justification for throttling access and routing sensitive queries to Opus 4.8.²

Taken together, the coverage portrays Fable 5 less as a simple product launch and more as a test case for a new AI playbook: open enough to excite investors and users, closed enough to reassure regulators and national security hawks – and controlled tightly by the company at the center.