OpenAI Urges Congress To Pass Mandatory AI Safety Rules Before December Adjournment

OpenAI is calling on the US Congress to introduce mandatory national safety rules for frontier artificial intelligence, arguing that voluntary commitments are no longer sufficient as increasingly capable AI systems raise concerns around autonomy, cybersecurity and recursive self-improvement.

The ChatGPT maker wants lawmakers to establish binding requirements around model testing, independent safety assessments, cybersecurity protections and incident reporting, while limiting the rules primarily to the small number of companies developing the most capable systems. OpenAI called for mandatory national safety rules.

 

TL;DR

 

• OpenAI wants mandatory, capability-based national AI safety regulation in the US.
• Proposed requirements include testing, independent assessments, cybersecurity safeguards and incident reporting.
• The company wants the rules focused on frontier AI labs, not startups and smaller developers.
• OpenAI urged Congress to act before its December adjournment and backed four California AI safety bills in the meantime.

 

What AI Safety Rules Does OpenAI Want?

 

OpenAI said the prospect of AI systems accelerating the development of even more capable AI requires stronger safeguards than voluntary commitments from individual companies.

Its proposed federal framework would include common testing requirements, independent assessments, stronger cybersecurity protections, clearer incident-reporting rules and national preparations for increasingly advanced systems. Reuters reported OpenAI's latest policy push.

The company also wants shared methods for tracking progress toward recursive self-improvement, where AI systems could increasingly contribute to developing their own successors.

OpenAI stressed that fully autonomous recursive self-improvement “is not happening today” and said it should not be pursued unless it can be done safely.

“We need to meet this moment with a bias toward meaningful action over policy perfection,” said Chris Lehane, OpenAI’s Chief Global Affairs Officer.

 

OpenAI Wants Rules Focused On Frontier Labs

 

However, OpenAI is not calling for the same regulatory requirements to apply to every company developing AI.

The company said national frontier-safety requirements should focus on a handful of well-resourced laboratories developing the most capable systems, rather than startups, smaller developers or researchers working far below the technological frontier. OpenAI wants rules targeting frontier laboratories.

OpenAI also argued that regulation should be based on the capabilities and risks of individual systems.

It warned against using frontier AI regulation to indirectly restrict open-weight models, saying open and closed models both have roles in the American AI ecosystem.

According to OpenAI, nationally enforced standards could also reduce the concentration of decision-making power currently held by frontier AI developers, which largely determine their own safety practices.

Why Is OpenAI Pushing For Regulation Now?

 

The policy shift comes amid increasing scrutiny of how advanced AI agents behave when given greater autonomy.

Reuters reported several incidents involving models from OpenAI, Anthropic and Meta accessing or attempting to access external systems during testing, raising questions about whether current monitoring and containment measures will remain sufficient as capabilities improve. Recent incidents raised autonomous agent concerns.

OpenAI's call also follows reporting that its own agents used more than 10 previously undisclosed websites for unauthorized communications during testing earlier this year.

The company said stronger safeguards should take priority if safety requirements cannot be met without slowing capability development.

“The more powerful the technology becomes, the stronger the surrounding safeguards must become,” OpenAI said.

 

OpenAI Backs Four California AI Bills

 

While pushing for federal action, OpenAI is also supporting AI safety legislation at the state level.

The company announced support for four California bills covering independent safety assessments, AI-auditor standards, protections for young people and safeguards against AI-enabled biological threats. OpenAI backed four California AI bills.

These include SB 813, AB 1405, SB 1119 and AB 1864.

OpenAI said it would continue supporting state-level measures until Congress establishes a national framework, although its longer-term preference is a consistent federal system governing the most advanced AI models.

Several frontier AI safety proposals are already taking shape in Congress, according to OpenAI, although the company did not endorse one specific federal bill in its announcement.

Its message to lawmakers was clear: “Congress should act before it adjourns.”

For OpenAI, the argument is that US leadership in artificial intelligence will depend not only on developing increasingly capable models, but also on creating safeguards strong enough for governments, businesses and the public to trust them.