News

Pentagon CTO Reaffirms Anthropic Remains Supply Chain Risk, Mythos Evaluation Doesn't Lift Ban

Pentagon CTO Reaffirms Anthropic Remains Supply Chain Risk, Mythos Evaluation Doesn't Lift Ban

Pentagon CTO Emil Michael recently countered reports of a warming relationship with AI firm Anthropic, stating unequivocally that despite the government's interest in Anthropic's Mythos model, the two entities are not "getting back together."

Speaking to CNBC, Michael reiterated that for the Department of Defense (DoD), Anthropic remains a supply chain risk unsuitable for its systems. He suggested this sentiment likely extends across the federal government but confirmed recent reports of some agencies evaluating Mythos for capability assessment, not operational deployment.

"The Mythos issue is a distinct national security consideration," Michael stated. "We must ensure our networks are hardened because that model possesses unique capabilities for identifying and patching cyber vulnerabilities."

Speculation about the administration's softening stance on Anthropic, following an acceptable use dispute with the Pentagon, had been circulating since mid-April after reports of the National Security Agency (NSA) using Mythos emerged. CEO Dario Amodei's White House visit last month further fueled speculation of broader government adoption, but Michael adamantly denies this.

"Anthropic represents a supply chain risk," Michael reiterated during an interview on Squawk Box. "From a national security standpoint, these factors must always be scrutinized. The NSA and Commerce Department evaluate all frontier models, including Chinese ones, to assess their cutting-edge capabilities."

In essence, the administration's official position is that any federal agency's use of Mythos is confined to model analysis, considered business as usual.

However, this doesn't preclude the government from exploring similarly capable emerging models, such as ChatGPT 5.5-Cyber. Reports indicate that AI leaders from multiple firms are scheduled to meet at the White House soon to discuss Mythos and the cybersecurity risks posed by advanced AI models.

Michael remarked on CNBC, "We view Mythos as the initial instance, but more will follow. The government intends to collaborate with all these companies in the coming year to first comprehend their capabilities, enabling us to address any issues across both private and public sectors."

↗ Read original source