OpenAI Calls on the US to Lead a Global Push for Binding AI Safety Standards

MakeCryippreferred on

• OpenAI called on the US to work with Congress on mandatory, capability-based AI safety regulation for frontier labs
• The company also urged development of compatible international standards for measuring AI capabilities and managing risk
• OpenAI warned of a closing window to establish safeguards before AI capabilities outpace the institutions meant to govern them

OpenAI published a policy statement calling on the US to lead a global push for binding AI safety standards, urging Congress to work toward mandatory, capability-based safety requirements for frontier AI labs rather than relying on voluntary industry commitments alone. The company’s call has drawn renewed attention this week amid separate discussions between the US and China touching on AI and trade.

The statement lays out four specific asks: national legislation establishing capability-based safety requirements, support for state-level measures including California bills addressing safety assessments and youth protection, voluntary industry standards for monitoring AI misalignment and incident reporting, and internationally compatible standards for measuring AI capabilities and determining when development should slow.

OpenAI’s global affairs lead Chris Lehane argued that stronger AI capabilities require stronger safety evidence, shared standards, and durable policy action while what the company calls the policy window remains open. The statement frames safety and progress as complementary rather than opposed, arguing that safety measures are what allow AI development to go further and benefit more people rather than acting as a constraint on it.

The company also took a notable position on the ongoing debate between open-weight and closed AI models, stating that the US needs both kinds of systems and cautioning against frontier safety policy becoming, in effect, open-weights policy under a different name. That framing pushes back on proposals that would impose safety requirements calibrated primarily around large closed models in ways that could disproportionately burden open-weight developers.

OpenAI’s call comes as multiple governments continue working through how to regulate frontier AI systems, with the EU, UK, and several US states each pursuing different regulatory approaches over the past two years. Whether OpenAI’s specific asks translate into actual legislation will depend on decisions by Congress and state legislatures that remain unresolved, and the company’s own statement acknowledges that establishing durable safeguards requires action before AI capabilities outpace the institutions responsible for governing them, not after.

The renewed attention to OpenAI’s policy position this week coincides with reported US-China discussions touching on AI and trade, a pairing that has drawn commentary questioning whether a US-led international standard-setting effort can realistically include Beijing’s participation given the broader competitive and security tensions between the two governments over advanced technology. OpenAI’s statement does not directly address how it expects China or other major AI-developing nations to engage with the compatible international standards it is calling for.