Anthropic Faces New Restrictions: U.S. Government Blocks Foreign Access to AI Models

The U.S. government has once again clashed with Anthropic by imposing a total ban on foreign access to the company's latest artificial intelligence models, "Fable 5" and "Mythos 5."

Anthropic announced on June 12 (local time) that "the U.S. government has issued export control guidelines, in accordance with national security directives, that completely prohibit all foreign nationals from accessing Fable 5 and Mythos 5."

Fable 5 and Mythos 5 are the official versions of the Mythos Preview, which previously triggered a global "Mythos Shock" by demonstrating overwhelming hacking capabilities.

Fable 5 is a version for the general public equipped with safety guardrails, while Mythos 5 is an unrestricted model provided selectively only to verified institutions through the global security consortium "Project Glasswing."

Anthropic explained that this measure blocks access not only from overseas but also for foreign nationals residing within the United States, including Anthropic's own foreign employees.

Anthropic stated that the letter sent by the U.S. government did not specify the exact nature of the national security concerns. However, the U.S. online media outlet Axios reported that the U.S. Department of Commerce decided on this measure due to security risks, following claims that a company had breached the security barriers of Mythos.

Anthropic argued, "The U.S. government appears to believe that users can bypass Fable 5's safety guardrails through methods known as 'jailbreaking.' We believe this stems from a misunderstanding, as the 'jailbreaking' methods identified by the government are widely used in other models, including GPT, and are functions used daily by security professionals."

While the company has suspended Fable 5 and Mythos 5 services for all customers, it pushed back, stating, "The government should have the authority to block the distribution of unsafe AI through transparent, fair, and technically grounded legal processes, but this measure does not align with those principles."

Previously, Anthropic also clashed with the U.S. Department of Defense over the usage scope of its AI model, "Claude."

Anthropic demanded that Claude not be used for domestic surveillance or fully autonomous lethal weapons. The Department of Defense rejected this and designated Anthropic as the first U.S. company to be labeled a "supply chain risk," leading to an ongoing legal battle between the two sides since March.

Reported by Kim Jin-u | Video by Na Hong-hee | Graphics by Lee Jung-ju | Produced by SBS Digital News
※ Please note: This article was translated by AI and may contain errors.