Anthropic’s announcement that the Trump administration had ordered the company to cut off foreign access to Fable 5 and Mythos 5 sent shockwaves across the tech industry, prompting the company to completely cut off access to these AI models.
Here are some reactions on social media to the latest skirmish in hostilities between the White House and Antropics.
Dean W. Ball Stephanie Augello/Getty Images for WIRED
Dean W. Ball, Senior Fellow, American Innovation Foundation
“If this is true, it’s just incomprehensible. We have an administration that is willing to do this. should We export advanced AI chips to China, but does China also want to ban the UK (and every other non-American on the planet) from using our best models? I have no words.
“I don’t know if this is a legal thing specifically for humanity or if it’s extreme national security high-handedness. Either way, it’s just a cartoon.”
Peter Girnus, Senior Threat Researcher, Zero Day Initiative
“Two things are true at the same time.
“First, Anthropic spent months marketing Mythos as too dangerous to release. Sam Altman said this was ‘incredible marketing to say we built a bomb.'” We’ve officially agreed that it’s a bomb. If you describe your product as munitions in every press release, the government will eventually take your word for it. They wrote the legal predicate themselves and called it a brand.”
“Second, we’ve done this experiment before. In the ’90s, the government classified encryption as munitions under the ITAR. Activists defeated encryption by printing the PGP source code as a book, because books are protected speech and floppy disks are arms exports. A T-shirt with three lines of RSA Perl was legally munitions. Control broke down because math doesn’t stop at customs.”
“The new problem is the ‘deemed export’ rule. Showing controlled technology to foreigners in the United States is considered exporting it abroad. So Anthropic’s foreign employees are now barred from the models they make. The ammunition is in the building, and the people who made it are not allowed to see it.”
Marc Andreessen, Partner at Andreessen Horowitz
Chris McGuire, Senior Fellow for China and Emerging Technologies, Council on Foreign Relations
“I actually think that targeted export controls on access to models are prudent. However, imposing blanket controls on a single model for all countries without any warning is highly questionable. Imposing similarly broad deemed export controls that also restrict foreign access is completely unreasonable and would clearly result in the model being removed from circulation, as just happened.
“Export controls are an important tool, a very powerful tool. Used correctly, they have the potential to significantly expand America’s lead in AI. Used incorrectly, they could stifle AI development. The Department of Commerce’s export control strategy is completely disjointed and sabotaging.” It is sending powerful AI chips to China without implementing regulations to prevent smuggling from China, creating massive loopholes that allow AI chips to be sent there, and preventing U.S. AI companies from releasing their own models.”
“This must stop. We urgently need a smart export control strategy that applies strong export controls that deny adversaries access to advanced technology while favoring U.S. companies. The Department of Commerce and BIS are consistently doing the opposite. If BIS doesn’t understand how to use its authority and the impact of its actions, it needs to find new talent who can actually implement a competent export control strategy. The current strategy is inconsistent and self-defeating.”
Matthew Pines, CEO of Physical Superintelligence
“This is going to send shockwaves through all the labs and neolabs… U.S. export control laws operate on a strict liability standard… They are very sharp knives…”
Dan Schipper, CEO of Every, a media and AI software company Alex Broadway/Web Summit Sports File (via Getty Images)
Dan Schipper, Every CEO
“My current view on this situation is that they will lift the ban within a few days and as a result, demand for fables will increase.”
“But this kind of thing is very disruptive and distracting for people in the company. The only comparable scenario I can remember is the firing of Sam Altman, which was resolved relatively quickly. I think that disrupted their momentum for a while, even if things got back to normal.”
“I hope we get a good result here!”
Josh Pigford Founder of Bearmetrics
“Anthropic has not profited from any hyperbole over the past six to 12 months. But I also assure you that this has nothing to do with national security.”
Peter Barnett, Researcher, Machine Intelligence Institute
“If it’s true that the USG restricted the export of Fable because another company said it could be jailbroken, we could end up with a situation where AI companies red-team each other’s models before the USG allows them to be deployed.”
Ryan Brewer, member of the OpenAI technical staff
“If the U.S. government continues in this direction, we will eventually only have access to frontier information in a small number of buildings in the Bay Area. That’s unfortunate.”
Ketan Ramakrishnan, Professor of Law, Yale University
“The federal government intends to aggressively regulate AI developers. The questions are whether this regulation will be done wisely, whether Congress and public deliberation will have a role, or whether it will simply be an opaque administrative action.”