Anthropic fails to convince Trump: Mythos 5 remains under embargo

Anthropic Faces Washington's Intransigence

Anthropic hoped to persuade Washington to ease the restrictions surrounding Mythos 5 and Fable 5, but Trump remained unyielding. Despite urgent discussions, U.S. restrictions remain in place. The latest AI model remains inaccessible internationally, caught in a standoff that extends far beyond mere technicalities.

On Monday, Anthropic met with several officials from the U.S. administration in an attempt to resolve the situation surrounding Mythos 5 and Fable 5. However, in the face of Trump, the company received no immediate concessions. Its most advanced model thus remains offline globally.

Mythos 5 vs. Trump: A Meeting, But No Turnaround!

Anthropic arrived in Washington with a mission: to demonstrate that its safeguards are robust enough to allow the return of its latest models. The table was filled with numerous security specialists from the company. Their goal was to explain why the vulnerabilities discovered in their latest model did not, in their view, pose a major threat.

However, the Trump administration remained steadfast regarding Mythos 5 and Fable 5. The U.S. government believes that certain flaws could allow for the circumvention of the model's safeguards, a scenario deemed concerning enough to maintain export restrictions. In other words, despite several hours of discussions, no one changed their position. The meetings mainly highlighted that the two sides do not exactly speak the same language.

Not Available Yet!

While awaiting a potential evolution, Mythos 5 remains blocked by the restrictions imposed under Trump, just like Claude Fable 5. This decision is already starting to have effects. Some developers are turning to competing solutions. Several industry experts are also worried about the consequences for American innovation. An open letter signed by several dozen researchers and industry leaders is calling for the swift lifting of restrictions.

According to them, these controls risk primarily slowing down American companies without preventing foreign competitors from advancing. However, for the administration, security remains the top priority. Yet, an interim solution is reportedly being considered. This would involve limiting access to only users located in the United States until Anthropic provides more technical assurances. However, it should be noted that no decision has been announced.

It's No Longer Just About Mythos 5

The issue is that this dispute does not solely concern Mythos 5. U.S. authorities are now seeking to regulate the most powerful AI models before their global release. The idea is to prevent them from being used to facilitate cyberattacks or other sensitive uses.

Anthropic, for its part, believes that the identified problems are limited and can be quickly corrected. The company also points out that other publicly accessible models have comparable capabilities. However, it is difficult to convince Washington when cybersecurity is part of the equation. In this domain, the precautionary principle seems to be taking precedence. Better to block first and discuss later. A philosophy that is certainly not universally accepted.

In any case, the standoff is far from over. And if there is one lesson to be learned, it is that the most advanced AI models are no longer just being developed in laboratories. They are now also being decided in the corridors of power.