White House Seeks Preapproval for Frontier AI Releases

According to Politico, the White House is considering an executive order that would create a government review process for advanced AI models, including testing and approval before public release. Politico reports that administration officials have met with tech companies about Mythos and similar models. Politico also quotes Kevin Hassett saying the proposal could provide "a clear roadmap to everybody about ... how future AIs that also potentially create vulnerabilities should go through a process so that they're released to the wild after they've been proven safe, just like an FDA drug." Reporting by Zvi on Substack and a related podcast states that the White House prevented Anthropic from expanding access to Mythos as part of Project Glasswing. FIRE.org and other commentators describe ongoing discussions about a Department of Commerce or interagency review that would test models for cybersecurity and national-security risks prior to broader deployment.

Industry observers note that formal pre-release testing for high-capability models presents hard technical trade-offs. Comprehensive security evaluation requires adversarial red-teaming at scale, reproducible benchmarks for misuse, and realistic operational testing environments, all of which are resource-intensive and often incomplete even inside major labs. Observers also point out that gating model access reduces the number of independent reviewers, which can limit discovery of novel failure modes that emerge only under broad usage.

Industry and civil-liberties commentators frame this debate as a tension between two governance philosophies: ex-post liability and ex-ante approval. Politico documents resistance from free-market think tanks and industry-aligned advisers; FIRE.org highlights free-speech and censorship risks if informal review pressures model behavior. Editorial observers note that ad-hoc or informal decisionmaking, such as putatively blocking access without published criteria, raises concerns about transparency and fairness in who gains early access to defensive capabilities.

• •Whether the White House issues an executive order or establishes a formal interagency review, and the legal authorities it cites.
• •Which agencies (for example, the Department of Commerce) are tasked with testing and what standards or metrics they publish.
• •Whether review processes are time-limited, publishable, or conducted under nondisclosure, and how that affects external security research.
• •Industry and civil-society responses, including legal challenges, congressional scrutiny, or coordinated disclosure frameworks.

Practitioners should monitor published procedures and metrics for any formal review regime, and anticipate increased emphasis on documented red-team results, risk classification, reproducible exploit demonstrations, and controls around staged access. Industry-pattern observers note that creating robust, transparent evaluation pipelines is technically feasible but will require investment in tooling, third-party testing capacity, and governance processes to avoid concentrating oversight benefits among well-connected actors.