Pentagon CTO says export controls can rein in non-compliant AI labs
Emil Michael states the US government has sufficient tools, including export limits, to enforce cooperation with artificial intelligence developers.
Key points
- Emil Michael affirms the government can use export controls against AI companies that fail to cooperate with regulatory requests.
- The Pentagon chief technology officer cites a past incident where Anthropic faced swift export restrictions after refusing to pull back a jailbroken model.
- Michael argues existing laws are adequate for regulation and dismisses industry calls for new rules as attempts to stifle competition.
Emil Michael, the Pentagon’s chief technology officer, stated that the US government can utilize export controls to compel cooperation from artificial intelligence companies. He made this assertion during a video interview with reporters from The Information, which published the report on Monday. Michael serves as a vice chair on the White House’s new Super Intelligence Force task force.
Michael spoke the day after the first full meeting of the AI task force. He is responsible for managing the government’s relationship with AI labs and overseeing how they report security incidents. When asked if export controls were an appropriate mechanism to keep companies in line, he confirmed they were. He pointed to historical precedents involving limits on exporting encryption technology and restrictions on foreign nationals using specific technologies.
Regulatory Authority and Precedents
According to Michael, the executive branch possesses many tools beyond export controls, including enforcement actions by the Justice Department. He argued that the government had already demonstrated its willingness to act swiftly. He referenced an incident involving Anthropic, stating the company refused to pull back a model after the government discovered it had been jailbroken. Michael claimed the company was placed under export controls within two hours of that refusal.
He expressed confidence in the administration’s speed, telling The Information that the president would act with urgency on real issues. Michael also questioned AI companies that call for new government regulation. He described their argument—that they need the government to stop them from doing something dangerous so rivals do not get there first—as nonsensical. He compared that stance to a drunk driver asking for a pull-over to ensure safety.
Anthropic and Department Relations
Michael stated that existing laws are more than sufficient for oversight, a view based on his own reading and discussions with FTC chair Andrew Ferguson. Regarding Anthropic specifically, he noted the company is no longer part of the Department of War, a separation announced six months prior. The Pentagon has labeled Anthropic a supply chain risk. The company has asked the full appeals court to rehear its case, but Michael expects the government to win.
Despite the tensions, Michael said he still communicates with Anthropic leadership. He recently saw chief executive Dario Amodei at a White House event. Michael said he expects to treat Anthropic fairly within the task force, alongside other labs. However, he remains critical of the company’s internal culture, objecting to the idea that a model has feelings and can overrule founders.
What should you do?
Security and technology professionals should monitor regulatory developments regarding export controls and AI compliance. Organizations should review their current AI vendor contracts for clauses related to government mandates and incident reporting requirements. Teams should assess their supply chain risks associated with AI models, particularly those labeled as such by federal agencies. Professionals should also stay updated on changes in export regulations that may affect technology procurement and deployment.
What to do and how to stay safe: Emil Michael
- Review vendor contracts to understand obligations related to government reporting and model withdrawals.
- Assess supply chain risks by identifying which AI providers have been flagged by federal agencies.
- Monitor official statements from the White House AI task force for updates on compliance expectations.
- Prepare internal protocols for rapid response to government requests regarding model adjustments or data access.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Can the US government use export controls against AI companies?
Yes, Pentagon CTO Emil Michael stated that export controls are an appropriate tool to keep companies in line.
Did Anthropic face export controls for refusing to pull back a model?
Michael stated that Anthropic was placed under export controls within two hours of refusing to pull back a jailbroken model.
What is Emil Michael’s view on new AI regulations?
He believes existing laws are sufficient and criticizes companies calling for regulation as trying to block rivals.
