/ THE SHORT ANSWER
OpenAI says it supports mandatory national safety requirements tied to model capability, continued state action until federal rules exist, independent safety assessments, AI-auditor standards, youth protections, and safeguards against AI-enabled biological threats.
- 01OpenAI is advocating mandatory capability-based national AI safety requirements.
- 02The company says it will continue supporting state legislation while federal action is absent.
- 03OpenAI identified four California bills it supports.
- 04The announcement is a policy position, not an enacted rule.
/ dotSuper point of view
When a frontier-model provider asks for mandatory safety rules, companies should prepare for capability-triggered evidence, independent assessment, and governance obligations rather than relying on voluntary policy alone.
What OpenAI is advocating
OpenAI says it wants mandatory national AI safety requirements based on model capability. The company also says it will support state legislation while Congress has not established a national framework.
OpenAI specifically named four California proposals covering independent safety assessments, AI-auditor standards, protections for young people, and safeguards against AI-enabled biological threats.
- Capability-based national requirements
- Independent safety assessment infrastructure
- Standards for AI auditors
- Youth and biological-risk safeguards
Why capability thresholds matter
A capability-based system applies stronger obligations when a model can create higher-consequence outcomes, rather than regulating every AI application identically. This can focus oversight but requires reliable tests and clear thresholds.
OpenAI’s own Astra safety overview reports Critical cybersecurity capability under its Preparedness Framework. That connection illustrates why evaluation evidence and deployment controls could become regulatory inputs.
- Define which capabilities trigger additional duties
- Specify who can test and verify results
- Address model changes after initial assessment
- Connect provider evidence to deployer controls
What enterprise buyers should prepare
Even when regulation targets frontier developers, enterprise customers may need evidence about model versions, evaluations, data handling, monitoring, incidents, and restrictions. Procurement should make that evidence retrievable rather than depending on marketing claims.
Organizations also need their own use-case governance. A capable model used in a low-risk drafting workflow is different from the same model controlling infrastructure, executing transactions, or advising on safety-critical decisions.
- Maintain a model and use-case inventory
- Record capability, access, and data boundaries
- Require incident and material-change notifications
- Escalate high-impact actions to accountable people
What this page cannot conclude
- 01The announcement states OpenAI’s policy position and preferred regulatory direction.
- 02The cited California bills may change during the legislative process.
- 03Final federal or state rules may differ materially from the company’s proposal.
Sources
- 01The AI policy window is open. We need to act.OpenAI · accessed Sep 10, 2026
- 02Safety overview: GPT-6 AstraOpenAI · accessed Sep 10, 2026
Our editorial standard · Found an error? Send a correction with its source.
/ CITE OR SHARE THIS GUIDE
Make the evidence easy to verify.
When you reference this guide, link to its canonical URL. That gives readers one stable place for the evidence, limitations and future updates.
dotSuper Research Desk. (September 10, 2026). OpenAI Calls for Mandatory AI Safety Rules: What the Policy Shift Means. dotSuper. https://dotsuper.net/feeds/daily-briefing/2026-09-10-openai-mandatory-ai-safety-rules-policy