OpenAI Backs US Bills to Curb AI Risks From Biological Weapons and Pushes Independent...

OpenAI is backing a series of bipartisan bills in the U.S. Congress aimed at preventing sophisticated artificial intelligence models from being used to accelerate the development of biological weapons and synthetic viruses, adding to the company's growing push for mandatory federal AI safety requirements.

The ChatGPT maker said Tuesday that it supports the Web of Biological Data Act, the AI-Ready Bio-Data Standards Act and the Scale Biology Act. OpenAI is also backing a key provision of the FRONTIER Act that would require leading AI companies to embed independent evaluators to assess the safety of their models.

The endorsements come as concerns over advanced AI have intensified inside the technology industry, including among executives and researchers at companies developing frontier models. OpenAI, Anthropic and other leading AI labs have recently debated whether the pace of development should be deliberately moderated as models become more capable.

OpenAI's support for congressional legislation is significant because the company has held that voluntary commitments by AI developers are not sufficient. Last week, it called for mandatory national AI safety requirements in the United States, proposing capability-based rules covering areas such as independent assessments, cybersecurity, and incident reporting.

The latest measures would extend that approach into biological security, where advances in AI and biotechnology are creating both opportunities for scientific research and concerns about misuse.

AI and Biology Become a Policy Battleground

The AI-Ready Bio-Data Standards Act would direct the National Institute of Standards and Technology to develop standards and frameworks for making biological data suitable for use in AI models. It would also establish minimum requirements for certain federally funded research so that the resulting biological data is AI-ready.

That may appear primarily focused on improving scientific infrastructure, but the underlying policy debate is broader. AI systems trained on sophisticated biological information could potentially accelerate drug discovery and other legitimate research while also increasing the ability of users to investigate biological threats.

OpenAI has already classified some of its advanced systems as presenting high biological capability and has introduced additional safeguards and external testing around biological misuse. The company says it has worked with U.S. and British government testing organizations and other external experts to evaluate these risks.

The other bills supported by OpenAI are part of the same effort to build a more formal infrastructure around AI-enabled biological research.

For the company, the issue is one of governance as much as model design. OpenAI's own Frontier Governance Framework covers risks including cyber offense, chemical, biological, radiological and nuclear threats, harmful manipulation and loss of control, while also setting out procedures for risk assessment, incident response and external expert input.

The proposed congressional measures would move at least some of those principles outside the companies themselves and into a federal framework. OpenAI has previously relied heavily on internal safety processes and voluntary third-party testing. Its new federal lobbying position would give independent evaluators a more formal role in assessing the most capable models.

Independent Evaluators Would Change the Safety Model

Under the provision OpenAI supports in the FRONTIER Act, leading AI companies would be required to embed independent evaluators to assess model safety.

OpenAI has already argued that outside assessments are necessary because internal testing can suffer from blind spots. Its own published guidance says third-party evaluations can provide additional evidence about frontier models' capabilities and the effectiveness of their safeguards.

The proposed requirement would nevertheless represent a significant change in the relationship between AI companies and regulators. Instead of companies largely determining how their most capable systems should be tested and how much information should be disclosed, independent evaluators would have a formal role in scrutinizing them.

That is becoming more relevant as AI models move beyond conventional chatbot functions. Modern frontier systems can use tools, maintain information across multiple steps, and operate within larger workflows, making their safety dependent not only on the model itself but also on the environments and permissions in which it operates.

OpenAI's endorsement also comes at a politically complicated moment for the industry.

Company executives have recently called for greater coordination over the development of advanced AI, with OpenAI CEO Sam Altman backing Anthropic CEO Dario Amodei's argument that frontier development should be paced so that safety measures can keep up. Other technology executives, however, have resisted calls for coordinated restrictions, arguing that companies should remain responsible for managing their own risks.

The political divide is sharper still. President Donald Trump has dismissed calls for tighter AI regulation while emphasizing the importance of maintaining U.S. technological leadership, particularly in competition with China.

OpenAI's position therefore places it in an unusual position: one of the world's most valuable AI companies is asking Congress to impose rules on an industry in which it is itself a major competitor.

The company says such regulation should be narrowly targeted at the most capable systems rather than applied broadly to smaller developers. Its policy proposals argue that frontier safety requirements should focus on well-resourced laboratories developing the most advanced models and that independent verification and transparency should replace fragmented private governance.

The biological-security legislation fits that framework. Rather than attempting to regulate every use of AI, the measures target specific infrastructure, data standards, and evaluation mechanisms that could become important as AI capabilities advance.

The unresolved question is whether Congress can turn those principles into enforceable rules quickly enough to keep pace with model development. OpenAI is now explicitly arguing that it should. The policy challenge is to preserve AI's potential to accelerate scientific discovery while ensuring that the same technology does not make dangerous biological knowledge easier to acquire, operationalize, or misuse.