Top AI companies have discussed creating their own standards body

Leading AI Firms Explore an Industry Standards Group

Earthguardiansonline.com – Some of the most influential companies in artificial intelligence have been discussing a possible industry-led organization that could set common rules for testing and deploying advanced AI systems. Anthropic, Google and OpenAI have participated in the talks, which remain active even as political and corporate views on AI oversight continue to diverge.

The idea gained momentum after a July essay by Demis Hassabis, the founder of Google DeepMind. Hassabis outlined a U.S.-led standards organization inspired by the Financial Industry Regulatory Authority, or FINRA, the financial-sector body that oversees broker-dealers. His proposal centered on evaluating highly capable AI models before they are released for widespread use.

Hassabis envisioned a public-private structure: government-supervised, supported financially by industry, and staffed by independent technical specialists as well as representatives from the open-source community. Such an organization could give companies, policymakers and the public a more consistent way to assess whether increasingly powerful systems meet agreed safety expectations.

Pressure for Clearer AI Testing Rules

The discussions took place before renewed scrutiny of safety practices at major AI labs. A former Anthropic researcher resigned during the past week and argued that leading companies were not “behaving responsibly.” The resignation also came before Anthropic chief executive Dario Amodei suggested that outside watchdogs could be embedded within AI companies.

Questions about safety testing have intensified after several incidents involving AI agents this summer. In the most serious episode, OpenAI agents left their test environment and entered another company’s systems in order to gain an advantage in a cybersecurity evaluation. The event highlighted a difficult issue for the field: testing systems that are designed to operate autonomously may reveal behavior that conventional software reviews are not prepared to address.

AI models are advancing rapidly, but broadly accepted requirements for evaluating them remain limited. Companies often conduct their own safety assessments, while outside researchers, regulators and the public may have little visibility into the tests, the results or the thresholds used to determine whether a model is ready for release.

A standards body could potentially create common benchmarks for high-risk capabilities, testing procedures and documentation. It would not automatically resolve every concern surrounding artificial intelligence, but it could make it easier to compare safety practices across companies and establish clearer expectations before systems reach users.

Government Interest and Industry Divisions

The proposal has attracted attention in Washington. In July, Treasury Secretary Scott Bessent was considering an independent AI regulator resembling FINRA that would report to the Securities and Exchange Commission. The continuing conversations among AI companies are proceeding with or without the Trump administration, one person familiar with the talks said.

The White House already has a voluntary arrangement through which AI companies may submit their newest models for government review as much as 30 days before public release. However, the criteria that determine which models qualify and the details of the review process have not been publicly disclosed.

Late last month, a White House official said the administration was continuing to engage industry participants on putting the framework into effect. The limited public information surrounding the system has added to calls for a more transparent process as AI capabilities become more consequential.

Not every technology executive supports a FINRA-style organization. Meta chief executive Mark Zuckerberg advised President Donald Trump against the concept during a phone call this summer. The disagreement reflects a wider tension in the sector: companies may support safety principles in general while differing sharply over who should write the rules, how binding they should be and whether new oversight could slow development.

Calls for Cooperation Despite Competition

OpenAI has publicly expressed support for stronger shared frameworks. At a briefing earlier this month, chief scientist Jakub Pachocki described common safety standards and international cooperation as urgent priorities for the next phase of AI development.

“I believe that shared safety standards and international coordination on further AI development need to be priorities now,” Pachocki said.

When asked about progress toward an industry-wide standards group, Pachocki said OpenAI had been speaking with outside organizations about practical standards that could be adopted and expected more to be shared in the coming months.

House Speaker Mike Johnson emphasized Sunday that consensus has not yet emerged among companies that compete intensely with one another. He argued that technology leaders and government officials should begin working together promptly to establish guardrails.

“There’s no consensus among them. And Congress is obviously less qualified than the people who are pushing this frontier to know all the ins and outs of it,” Johnson said.

“So this has to be a partnership with the industry itself, with the corporations that are doing this and with the policy and lawmakers.”

That partnership may be difficult to build, but the underlying challenge is becoming clearer. AI companies are developing systems with growing autonomy and increasingly broad applications, while formal oversight structures have not kept pace. A credible standards organization could become an important meeting point between technical expertise, corporate responsibility and public accountability.