Anthropic Logo [Reuters=Yonhap News]

The Trump administration in the United States had considered export controls after identifying an issue where Anthropic allowed a company with suspected ties to China to have early access to its latest artificial intelligence (AI) models, 'Mythos 5' and 'Fable 5', before their official launch, reported the U.S. daily Washington Post (WP) on the 15th (local time).

According to White House officials cited by WP, Anthropic submitted a list of 111 organizations that were granted early access to its latest AI model, 'Mythos', to the Trump administration a few weeks ago, and the administration reviewed and approved the list.

However, Anthropic later revealed that an additional approximately 50 organizations had received access. When the company did not disclose the identities of the newly added organizations for several days, the administration began considering export controls.

Officials stated that the Trump administration eventually discovered that a South Korean telecommunications company, suspected of being linked to China, was included in the list provided by Anthropic.

In the pre-launch phase of 'Mythos 5' and 'Fable 5', Anthropic had allowed some overseas institutions and companies, including those in the United States, to use Mythos in advance through Project Glasswing.

Project Glasswing is a global cybersecurity collaboration launched by Anthropic when it announced the development of Mythos. It operates by preemptively providing the model to verified companies and institutions to defend against software vulnerabilities, out of concern that Mythos could be misused by malicious hackers.

A source well-acquainted with Anthropic's situation told WP that Anthropic promptly revoked the Mythos access for the South Korean company.

However, this incident is known to have eroded the trust of Trump administration officials in Anthropic's ability to protect sensitive technology.

WP reported, "Weeks before forcing the company to halt services, the Trump administration had already been reviewing export controls after a controversy arose regarding the sharing of technology with a company suspected of ties to China."

The U.S. government announced export control guidelines on the 12th to prevent foreign institutions and individuals from accessing Anthropic's top-tier AI models, 'Mythos 5' and 'Fable 5'.

In response, Anthropic immediately suspended services for 'Mythos 5' and 'Fable 5' for all users to comply with the guidelines.