Tensions between Anthropic and the US administration have reached a breaking point. In April, Anthropic announced it had built Mythos, an AI model so adept at coding that it could pose a global cybersecurity threat. The company granted access to a small group of experts for evaluation, then released a modified, safer version called Fable on June 9. The federal government swiftly responded by imposing export controls, declaring Fable a national security risk. Anthropic revoked access to both models within hours.
Export Ban Hits Anthropic Unexpectedly
The government intervention was rapid and severe. Amazon CEO Andy Jassy informed officials of the danger posed by Fable, despite Amazon being both an investor in Anthropic and a developer of competing AI models. Many experts question the legal basis of the ban, as offering cloud access to a model may not constitute exporting. However, the decision has already triggered ripple effects.
Sponsored Protocol
Cybersecurity Experts Warn of Increased Vulnerability
Paradoxically, the ban may increase the country's vulnerability to cyberattacks. In an open letter to the government, leading cybersecurity experts emphasized that access to Anthropic's models helped researchers prepare defenses. Other equally capable models remain widely available, making the ban ineffective if not harmful. Applying the concept of nonproliferation to software is proving challenging.
European and Chinese Ripple Effects
The decision is driving many to avoid reliance on US AI companies. French politician Bruno Retailleau called it a wake-up call for Europe to develop independent AI. However, China offers highly capable, cheap open-source models that can be downloaded without restrictions. Shares of Chinese startup Zhipu have surged, indicating companies may prefer Chinese models to avoid government interference. This could lead to a new crackdown: the US government might declare the use of Chinese models a national security threat.
Sponsored Protocol
What Lies Ahead for AI Regulation
Anthropic's previous feud with the government over Pentagon usage of its models led to new legislative proposals. Now, the White House has labeled one of the most valuable AI startups a national security risk, first in spring and again in summer. Lawmakers are still far from a clear regulatory framework, but with each drastic action, pressure for federal AI regulation mounts. Polls show most Americans want more rules. The future of AI remains uncertain as the administration's stance shifts with the wind.
Sponsored Protocol
For context on digital infrastructure, read our article on Inside Rogfast: The World's Deepest and Longest Subsea Tunnel and Flexible Data Centers for AI. For more on export controls, visit Wikipedia.
