The Trump administration has lifted export controls on Anthropic's powerful Fable and Mythos AI models. In a letter to Anthropic co-founder Tom Brown, Howard Lutnick of the Commerce Department announced that a license is no longer required for these advanced models.

Previously limited to select companies and government agencies, the new policy means broader access. However, Anthropic has agreed to bolster security measures, particularly to prevent 'jailbreaks'—ways users might bypass safety restrictions. This cooperation comes as Dario Amodei was recently replaced in meetings with Brown, whom officials now prefer.

The move follows a contentious period during which Anthropic argued its models were secure enough despite the administration's concerns about potential risks to cybersecurity. Now, with the new agreement, the focus shifts to practical security solutions rather than theoretical debates.

These developments in AI policy could have far-reaching implications for how governments and tech companies navigate the complex landscape of AI research and deployment. As ever, the future of AI is written in code—but perhaps also in compromise.