In a low‑key ceremony on Tuesday, President Donald Trump signed an executive order that reshapes the federal approach to regulating cutting‑edge artificial intelligence. Called “Promoting Advanced Artificial Intelligence Innovation and Security,” the order replaces a draft that would have forced developers to submit models for a 90‑day mandatory review with a framework that asks firms to share their frontier AI systems voluntarily, up to 30 days before a public launch.
The new order’s centerpiece is a voluntary pre‑release review mechanism. Companies can approach the government to determine whether a model qualifies as a “covered frontier model,” grant access for a limited window, and work with designated “trusted partners” for early testing. Participation is optional; firms that decline face no penalties. The shift from mandatory to voluntary reflects a decisive win for the pro‑industry faction inside the White House, which argued that heavy regulation would stifle U.S. innovation and put American firms at a disadvantage against Chinese rivals.
Beyond the review process, the order establishes an AI cybersecurity clearinghouse to be set up within 30 days. The clearinghouse will be coordinated by the Treasury Secretary, the National Cyber Director, the National Security Agency and the Cybersecurity and Infrastructure Security Agency. Its mandate is to scan AI software for vulnerabilities, validate findings and shepherd remediation and patch distribution. The move directly addresses the so‑called “Mythos” incident, where AI‑discovered flaws outpaced existing disclosure mechanisms.
Federal agencies are also tasked with developing benchmarks to evaluate the cybersecurity capabilities of AI models and strengthening the government’s own defenses against AI‑enabled threats. While the order mentions AI safety research, it offers fewer concrete directives than the earlier draft, leaving much of the work to future agency rulemaking.
Key provisions that were cut include the original 90‑day mandatory review period, a formal authority for the government to evaluate models, and stricter reporting requirements that mirrored elements of the Biden administration’s repealed AI order. Industry leaders, including Google, Microsoft and Elon Musk’s xAI, have already engaged in voluntary pre‑release testing with the Center for AI Safety Initiative (CAISI). Those companies may continue or expand cooperation, but firms that view government oversight as a commercial hindrance can simply opt out.
The European Union’s AI Act, set to be fully enforced in August, stands in stark contrast. The EU law imposes mandatory compliance, statutory authority and penalties for non‑compliance. By contrast, the Trump order builds institutional infrastructure—such as the clearinghouse and benchmark development—while relying on goodwill rather than legal compulsion.
Analysts note that the order’s impact will be measured by industry participation. If major developers continue to share models, the government could gain valuable insight into emerging risks and coordinate faster vulnerability patches. If companies choose to bypass the framework, the order may serve more as a political talking point than a regulatory lever.
For the administration, the quiet signing offers a policy document that can be cited in future debates about AI oversight, without provoking a public clash with a tech sector that counts many of its leaders among the administration’s most visible supporters. Whether a voluntary framework can keep pace with AI’s ability to uncover thousands of zero‑day vulnerabilities each month remains an open question.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory SEO vous aide à automatiser le contenu d'actualités pour votre site.