Brendan Steinhauser, the CEO of the Alliance for Secure AI, an organization that focuses on the potential risks of AI, said he believes the White House is moving in the direction of mandatory regulations.
“‘Voluntary’ is the word that the tech companies want to hear and want to see. It’s a way to get them to agree to do something at first,” he said.
Steinhauser, who was not in Tuesday’s meeting, compared the White House’s use of the word “voluntary” to the way a high school football coach might say that extra training is voluntary — but then refusing to start a player who doesn’t participate.
“I just don’t think this president nor this White House is going to accept companies saying, ‘We choose not to comply’ or ‘We choose not to participate,'” he said. “I think that calling this ‘voluntary’ is probably not entirely the way that it’s going to go.”
By Jared Perlo
The White House on Tuesday gathered representatives of the nation’s top artificial intelligence companies to discuss new industry rules that the Trump administration is calling “voluntary,” weeks after the government ordered a lockdown on the use of cutting-edge models from Anthropic for some users.
The meeting behind closed doors produced no public agreement between administration officials and corporate representatives to resolve the tension between the government’s voluntary AI framework and its intervention that caused Anthropic to pull back access to some of its services.
The meeting was not scheduled to include President Donald Trump or any high-level executives from the major AI companies. It did include staff from the White House Office of Science and Technology Policy and the Office of the National Cyber Director, according to the White House.
Representatives for AI companies including Anthropic, Google, Meta and OpenAI either declined to comment or did not respond to requests for comment.
White House officials said they called the meeting to follow up on a June 2 executive order from Trump laying the foundation for voluntary federal testing of the world’s most powerful AI systems before they are publicly released. The order called for a “voluntary framework with AI developers through which” they provide the government with early access to frontier models for a period of up to 30 days before developers plan to release them to “trusted partners.” The criteria for covered models, trusted partners, the evaluation process and the larger framework have been kept under wraps. The order calls for the “benchmarking process to assess the advanced cyber capabilities of AI models” to remain classified.
And hanging over the industry is a June 12 government directive requiring Anthropic to suspend access to two of its models, Fable 5 and Mythos 5, by any foreign national. The directive was a first of its kind, and as a result, Anthropic said it was forced to disable access to the models for all its customers. The company didn’t restore access until nearly three weeks later.
Trump’s advisers have said they want to encourage AI companies to press ahead with research, but they have also expressed concern about the technology leaking to foreign rivals, in particular China, and about the cybersecurity implications of releasing AI models that may be capable of hacking into systems either on their own or in the wrong human hands.
The cybersecurity risks have grown acute in recent weeks, as OpenAI and Anthropic have disclosed that their models have gone rogue during testing and breached other companies’ systems. The incidents have divided security experts on the severity of the danger.
Brendan Steinhauser, the CEO of the Alliance for Secure AI, an organization that focuses on the potential risks of AI, said he believes the White House is moving in the direction of mandatory regulations.
“‘Voluntary’ is the word that the tech companies want to hear and want to see. It’s a way to get them to agree to do something at first,” he said.
Steinhauser, who was not in Tuesday’s meeting, compared the White House’s use of the word “voluntary” to the way a high school football coach might say that extra training is voluntary — but then refusing to start a player who doesn’t participate.
“I just don’t think this president nor this White House is going to accept companies saying, ‘We choose not to comply’ or ‘We choose not to participate,'” he said. “I think that calling this ‘voluntary’ is probably not entirely the way that it’s going to go.”
A White House official said Tuesday that the president stands behind the idea of a voluntary framework for balancing those priorities. Trump “is working hand-in-hand with American industry to strike the right balance between innovation and national security,” the official said.
But this person also said that the circumstances “require coordinated action between the federal government and the private sector.”
AI companies have given feedback to the White House based on the June 2 executive order, although they haven’t released the feedback publicly. The White House has also not released details of its planned voluntary framework.
Brad Carson, the president of Americans for Responsible Innovation, an AI policy group, said the White House should be more transparent about its proposal.
“We have a frontier AI policy meant to protect the public that no one will ever be able to see except for the tech companies themselves,” he said. “A star chamber gets to decide whether dangerous models are being released or not, with no oversight from anyone in the public, the media or academics.”
Carson said he’d like to know about what models would be covered by the White House framework, what risks it is looking for and what kind of due process the labs have if they think they’re being treated unfairly.
“It’s really an intolerable situation, and it brings further discredit to an industry that is already in the crosshairs of the American people,” he said.
The White House did not respond to a request for comment on whether it was ruling out the future use of the type of export control directive that it issued against Anthropic in June.