Startups see AI export ban as test of US model power rules

The AI export ban reversal ended a weekslong block on Anthropic’s Fable model, but it opened a sharper fight over federal AI control.

The Trump administration said Tuesday it was lifting the restriction, according to the account provided. The decision mattered less for ending one dispute than for revealing a new question facing AI labs, startups and security officials: who decides when a frontier model is safe enough to release?

Fable reversal leaves doubts

Hours after the reversal, Quinn Slack, founder and chief executive of AI-agent startup Amp, addressed about 50 technology workers at a San Francisco rally called “Freedom of Intelligence.” Slack said the episode showed that Washington could restrict access to AI systems even after companies had built products around them.

“That is a dangerous precedent to have set,” Slack told the crowd, referring to the government’s role in deciding access to Fable. His concern reflects a wider industry fear that a temporary export block can become a template for future intervention.

The source account did not include the text of an official order, the exact export mechanism or the full criteria used to pause Fable. That limits what can be stated about the legal basis for the action, but the political signal was clear: model capability is now being treated as a national-security question.

Cyber tests redraw boundaries

The administration’s concern centered on whether models from Anthropic and OpenAI could help malicious actors discover previously unknown software flaws. If a model can accelerate vulnerability research, officials may view its release as more than a commercial launch.

That logic marks a turn from the earlier hands-off posture toward AI oversight described in the source material. Instead of relying mainly on company controls and voluntary safeguards, the government created what critics called a de facto approval process for powerful models.

Alex Stamos, chief product officer of AI security firm Corridor, warned that the episode could distort how AI labs test their own systems. “U.S. labs are getting the message that they should make sure that their models are never very good at cyber evaluations, lest they land in endless model purgatory,” Stamos wrote on X after the Fable restriction was lifted.

Will Rinehart, a senior fellow at the American Enterprise Institute, said the administration’s handling of Fable unsettled the market for frontier models. The effect, he said, “was to put the entire industry on a shaky foundation.”

Rinehart framed the core shift in sharper terms: “We have entered a world where a frontier model is considered ‘secure’ when the government says so.” That view captures the tension between federal security claims and the industry’s demand for predictable release rules.

Three paths for Fable

If Washington turns the Fable episode into a clear review system, the global macro effect could be a slower but more predictable rollout of advanced AI tools across borders. Anthropic would face added compliance work, while the wider sector would gain a playbook for cyber evaluations and export decisions.

If the process remains informal, uncertainty becomes the mechanism. Enterprise buyers may delay deployments, Anthropic could face pressure to prove model safety before customers commit, and rival labs may reduce public cyber testing to avoid regulatory attention.

If officials retreat from model-level controls after the backlash, AI companies would regain speed in product launches, but the cyber-risk debate would move back to company governance. The open questions are concrete: what capability triggers review, who evaluates the test results, and whether restrictions apply equally to Anthropic, OpenAI and future competitors.