Anthropic, Google, and OpenAI have discussed creating an industry standards body that could establish common standards for evaluating advanced AI models before they are widely deployed.
The conversations are ongoing and could continue even without participation from the Trump administration, according to CNN. No agreement has been reached on what the standards would require, but the effort could eventually give technology partners and enterprise customers a more consistent way to assess AI models before deployment.
Rival AI companies consider a shared standards body
According to CNN, the discussions were sparked by a July proposal from Google DeepMind co-founder and CEO Demis Hassabis. He called for a US-led standards body modeled on the Financial Industry Regulatory Authority, or FINRA.
The proposed organization would test advanced AI models before broad deployment. CNN reported that it could operate as a public-private partnership overseen by the government, funded by industry, and staffed by independent technical experts and representatives from the open-source community.
OpenAI has also been speaking with outside organizations about what specific standards could look like.
“I believe that shared safety standards and international coordination on further AI development need to be priorities now,” OpenAI chief scientist Jakub Pachocki told reporters earlier this month, per CNN.
Quartz noted that Pachocki said OpenAI has discussed potential concrete standards with external organizations and expects to have more to share in the coming months. OpenAI CEO Sam Altman has also publicly expressed support for collaborating with other AI companies on the effort.
The talks remain at an early stage. CNN said that the companies have not reached a consensus on what specific standards or guardrails should apply.
US government weighs a similar approach
The industry discussions overlap with a separate proposal considered by the Trump administration.
Bloomberg said in July that Treasury Secretary Scott Bessent helped develop a plan for an independent AI regulator that would evaluate the safety of advanced models with industry input. The proposed organization would report to the Securities and Exchange Commission and follow a structure similar to FINRA.
A June executive order directed federal agencies to design a voluntary framework through which developers could provide the government with access to designated frontier models for up to 30 days before releasing them to other trusted partners. The order does not establish mandatory government approval before release, and details of the benchmarking and designation process have not been made public.
Not every major technology company supports creating a national AI regulator. Meta CEO Mark Zuckerberg reportedly argued against the idea during a call with President Donald Trump this summer, calling the concept fundamentally flawed.
What common AI standards could mean for the channel
A common testing framework could give resellers, integrators, consultants, and other channel partners another way to compare AI products from competing vendors.
Partners already have to evaluate differences in model capabilities, security controls, deployment requirements, data handling, and governance before recommending AI platforms to customers. Agreed-upon testing standards could make those comparisons more consistent, particularly if major vendors begin publishing results under the same framework.
Enterprise customers could also begin asking whether AI products meet recognized standards in the same way they currently evaluate security certifications, compliance requirements, and vendor risk.
For MSPs, the development could be especially relevant as more customers turn to service providers for help selecting, deploying, and managing AI tools. Common standards could give MSPs another benchmark for deciding which models to support and for explaining the risks and differences between competing platforms.
Much still depends on whether Anthropic, Google, OpenAI, and other AI companies can agree on common standards. Channel partners do not need to change their model-selection processes yet, but they should watch whether the proposed body produces comparable test results, independent oversight, or recognized certifications that customers begin including in procurement requirements.
For another look at AI industry collaboration, see how OpenAI, Google, Meta, and Anthropic joined the White House to discuss cybersecurity testing for advanced AI models.