OpenAI has called on US Congress to introduce mandatory national safety requirements for the most advanced AI systems, arguing that policymakers have a limited opportunity to establish safeguards before AI capabilities outpace existing governance frameworks.
In a blog post published by chief global affairs officer Chris Lehane, the company said it is entering a “new chapter” in AI development that requires a corresponding shift in public policy, with stronger regulation, industry standards and international co-operation.
The company said it is seeking capability-based national legislation covering frontier AI models, including requirements for independent safety assessments, cybersecurity protections, incident reporting and national preparedness.
OpenAI also pledged to continue supporting state-level legislation until federal rules are introduced, announcing its backing for four California bills covering independent AI safety assessments, AI auditor standards, protections for children using AI companions and safeguards against AI-enabled biological threats.
OpenAI argued that AI development is reaching a point where technical safety work within individual companies is no longer sufficient on its own.
Lehane pointed to recent advances in AI-assisted research, saying that while fully autonomous recursive self-improvement is not yet a reality, AI systems are already helping accelerate the development of future models.
“The AI policy window is open, for now. We intend to use it,” he wrote. “That means acting with urgency, humility, and a willingness to adapt.
“It means supporting serious proposals that materially raise the safety bar, even when they are not exactly what we would have designed. And it means strengthening the framework as the technology evolves.”
OpenAI said it will continue to slow or halt the development or deployment of systems where it believes safety standards cannot be met, in line with its preparedness framework.
Alongside regulation, the company is calling for frontier AI developers to establish voluntary industry standards for monitoring advanced models, particularly for signs of misalignment, where systems pursue objectives in ways that diverge from human intent.
The proposed framework would also include clearer disclosure requirements for significant AI security incidents and broader reporting obligations for frontier model developers.
OpenAI said it hopes any industry-led standards will complement future legislation rather than replace it.
The company also renewed calls for internationally compatible AI governance, arguing that global standards will ultimately be needed for measuring frontier capabilities, managing risk and determining when development should slow or stop.
