The Trump administration has lifted export controls on Anthropic's Claude Fable 5 AI model, restoring customer access. This powerful model was previously pulled for security concerns, specifically its potential for misuse in cyberattacks or bioweapon development. Last week, limited access was restored to government-approved organizations after Anthropic implemented new safeguards. These safeguards reportedly block jailbreaking attempts with high success rates, though they may also flag benign requests. The Commerce Department confirmed its collaboration with Anthropic to ensure national security alignment and strengthen US AI leadership. This ad hoc regulatory approach highlights the evolving government role in evaluating frontier AI models. OpenAI has also adopted a similar limited release for its GPT-5.6 model to approved customers. The US government has an August deadline to establish standardized AI security risk benchmarks. Meanwhile, China is developing its own advanced AI models, posing increasing competition.