The U.S. government this week issued an export control directive that forces Anthropic to suspend access to its Fable 5 and Mythos 5 AI models for foreign nationals — and the company says it had to disable those models worldwide as a result. The move is already reshaping how investors think about exposure to centralized artificial intelligence. Over the same stretch, Bittensor (TAO), a decentralized AI network, surged nearly 30%.
Why the export control landed hard
Anthropic disagrees with the order. The company argues the vulnerability the government cited is minor and already present in other publicly available frontier models. But the directive didn’t leave room for negotiation. Anthropic had to kill access to two of its flagship products globally, not just for users in the targeted countries. For investors, the message is blunt: a single government order can turn off a centralized AI service overnight.
Bittensor’s jump
Grayscale Head of Research Zach Pandl highlighted decentralized AI as a major opportunity, calling Bittensor the protagonist in that shift. He compared Bittensor to Bitcoin — but for AI — aiming to give permissionless access to computing resources. Bittensor works through a distributed architecture of subnets that reward contributors with TAO tokens. The token has run up roughly 30% over the last five days, partly on the back of the Anthropic news and partly on growing interest in alternatives that can’t be shut down by a single regulator.
Anthropic’s legal headache
As if the export control wasn’t enough, Anthropic is also facing a class action lawsuit over the usage limits on its commercial Claude products. The complaint centers on the gap between what subscribers get. Max 5x, the suit claims, provides about 3.5 times more usage than the Pro tier, while Max 20x gives roughly 6 times Pro — and Max 20x is only about 1.7 times Max 5x. Users say the tiers don’t deliver the value they expected. The timing isn’t great for a company already under regulatory pressure.
Investors rethink centralized AI
The combination of a government-imposed switch-off and a user backlash is pushing some money toward models that don’t depend on a single corporate server. Decentralized networks like Bittensor spread compute across multiple nodes, making them harder to block or censor. The question that remains open is whether these networks can scale fast enough to meet enterprise demand — or whether the U.S. and other governments will eventually try to regulate them too.
