Anthropic [Reuters=Yonhap News file photo]

It has been revealed that the U.S. administration's decision to restrict foreign access to Anthropic's latest artificial intelligence (AI) models, such as 'Mythos 5' and 'Fable 5', was due to a tip-off from Amazon, an investor in Anthropic.

Andy Jassy, CEO of Amazon, informed government officials that their researchers had discovered that Anthropic's 'Fable 5' model could be used for cyberattacks, leading the government to issue a blocking directive, the U.S. daily Wall Street Journal (WSJ) reported on the 13th (local time), citing multiple sources.

Although Fable 5 was designed to block commands that could be misused for cyber warfare or biochemical weapons, Amazon researchers managed to bypass these restrictions by inputting a series of commands (prompts).

U.S. government security personnel verified Amazon's claims and concluded that the most direct way to address this risk was to prevent foreign governments, companies, and individuals from accessing the tool.

A senior White House official stated that President Donald Trump gave the final approval for these measures.

Amazon has been a key investor in Anthropic, having invested a total of $13 billion (approximately 20 trillion won) from 2023 to the present, and has also pledged an additional $20 billion in investment tied to Anthropic's future commercial milestones.

However, this year, Amazon has also been placing bets on OpenAI, committing to invest up to $50 billion (approximately 75 trillion won) in the company.

Amazon stated, "As a leading cloud provider supporting numerous private and public sector customers, it is not uncommon for the government to seek our advice on potential security risks," adding, "We do not share the specifics of such discussions."

The WSJ also suggested that the conflict between Anthropic and U.S. government agencies, including the Department of Defense, which is currently engaged in litigation, may have influenced this decision.

Adam Thierer, a senior researcher at the think tank R Street Institute, pointed out, "We are currently witnessing the increasing politicization of AI in this country and the concentration of control over advanced computing."

Kate Klonick, an associate professor at the Center for Strategic and International Studies (CSIS) and former official at the Department of Commerce's Bureau of Industry and Security, acknowledged the security concerns but analyzed, "The White House's antipathy towards Anthropic likely influenced this decision."

Conversely, a senior White House official explained that the measures are related to model safety and that the Department of Defense was not significantly involved.

On the 12th, the U.S. government announced export control guidelines that completely prohibit access to 'Fable 5' and 'Mythos 5' for all foreign nationals. In response, Anthropic temporarily suspended access for all users to these models to comply with the directive.

Anthropic stated that claims of the Fable 5 model being bypassable are a misunderstanding and that they are working to restore service.