A senior Trump administration official tells FOX Business that Anthropic’s “recklessness” in responding to issues with the company’s latest artificial intelligence release led to export controls.
The official says there are ongoing talks about fixing the vulnerabilities with the company’s Fable 5, but claims the administration’s interactions with Anthropic created a lack of trust. The official believes the experience damaged Anthropic’s relationship with the government over its unwillingness to address the concerns.
The export controls imposed by the Commerce Department forced the company to pull down its latest release. Anthropic received a letter from the Commerce Department saying its Fable 5 and Mythos 5 models were banned for use by any foreign national “inside or outside the United States,” due to national security concerns. A source familiar with calls to Anthropic says the company brushed off the federal government’s initial concerns. A source close to the company disputes that, adding it was not presented with any details and never refused to fix issues.
ANTHROPIC FILES CONFIDENTIALLY FOR IPO
After a flurry of phone calls on Friday, the official familiar with the outreach tells FOX Business the company claimed Anthropic CEO Dario Amodei was at a wellness retreat and could not be reached about the concerns. A source close to Anthropic denies there was a retreat, says the company’s executives were not hard to reach and claims they were in touch with the White House within 15 minutes.
Still, the senior administration official says Anthropic’s reaction to potential issues led to the export controls, adding the government says the company did not take the request to fix issues seriously enough, given the expectations set before the Fable 5 rollout.
The feeling within the Commerce Department, Treasury Department, and White House is that Anthropic could have avoided the drastic step of export controls by engaging on a deeper level.
Multiple sources say the government made it clear to Anthropic that this release would be the first test case for the new executive order signed regarding AI guardrails. A Washington, D.C., tech policy expert tells FOX Business this is “government messaging 101” – that if a company is leading the conversation on AI safety, then appears reluctant to address any safety concerns, the administration will take a tough stance.
AI LEADERS ARGUE SOFTWARE WILL ADAPT – NOT DIE – BUT VALUATIONS ARE STRETCHED
Administration officials say the concerns were realized last week when Amazon AI experts were able to orchestrate a “jailbreak” on Fable 5 after it was released on June 9 alongside Mythos 5, the latter of which was only released to a limited group. Anthropic said the cyberattacking abilities of the new models were constrained, but Amazon and five other companies testing the release found a workaround in Fable 5 that opened the full cyber abilities of Anthropic’s more advanced model.
“As a leading cloud provider that serves a large number of private and public sector customers, it’s not uncommon for governments to seek our counsel on potential security risks,” an Amazon spokesperson said in a statement. “When they occur, we don’t share the details of these discussions.”
Under its leadership, the Trump administration has taken an all-of-government approach to making sure new AI models meet national security standards. The export controls originated in the Commerce Department, but the Treasury Department is being increasingly looped in due to its knowledge of computer power and markets.
A former Trump administration official tells FOX Business that President Donald Trump is business-friendly to a point, adding that if a company is slow to address issues, the president will force it to make the necessary changes. The former administration official expressed concern that the export controls will not be easily removed if Anthropic continues to brush off the issues being raised at the highest levels of government.
AMERICA CAN’T COMPETE WITH CHINA IN AI WITHOUT THESE WORKERS, META’S PRESIDENT SAYS
In a statement on June 12, Anthropic said, “We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities. These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass.”
A source close to Anthropic says there were no details in the initial call with government officials, just a request to pull down the Fable 5 model within 90 minutes. The source adds that Amodei and other senior staff have been engaged from the beginning, there was no refusal to fix an issue and the company remains eager to resolve any concerns.
FOX Business has also learned Anthropic senior technical staff are in Washington, D.C., to meet with White House officials. Another source tells FOX Business virtual meetings have been held every day since the administration’s initial outreach about the vulnerabilities.
Read the full article here