Experts urge lifting Claude Fable 5 controls

Arbitrary export controls on cybersecurity-focused AI models are a self-inflicted wound that hamstrings domestic defenders while failing to stop adversaries who have access to equivalent global alternatives.

• –The distinction between offensive and defensive AI capabilities is often artificial; blocking tools that detect code vulnerability directly harms software security research and development.
• –Unilateral bans on specific frontier models are futile in a globalized ecosystem where open-weight alternatives and competitor models (like Kimi 2.7) offer similar capabilities.
• –Policymakers must transition from ad-hoc, panic-driven prohibitions to transparent, scientifically validated evaluation criteria developed in collaboration with the security industry.