Trump just found the worst way to regulate AI
TL;DR
Trump's administration forced Anthropic to take Fable offline through export controls after Amazon warned officials about a possible jailbreak. Fable is the heavily restricted public version of the stronger Mythos model. Anthropic says thousands of testing hours found no universal safeguard bypass. Vox argues the issue is not AI regulation itself, but a process with no clear standards, no congressional mandate, and little public cost-benefit analysis.
Nauti's Take
This is regulation by panic button: a cloud giant flags a risk, the government yanks the plug, and builders are left decoding the rules after the fact. Model teams need hard benchmarks and appeal paths, not export-control hammers that swing with political gravity.
Briefingshow
Frontier models need oversight, but this case shows how easily safety regulation can turn into executive discretion. If the US can pull access to a major model without clear public criteria or a transparent review, customers, researchers and foreign partners have a reason to hedge against American AI platforms.
Tweets
Sources