Google, OpenAI and Anthropic, three of the world’s leading artificial intelligence companies, have reportedly moved toward forming a new “Frontier AI Standards Agency” focused on developing common standards for testing and safety across advanced AI systems.
According to The Information, the three AI labs are reportedly exploring the creation of the agency and have approached Sriram Krishnan, a well-known Silicon Valley policy figure, about serving as its CEO.
The reported move would represent an unusually high level of coordination between companies that are otherwise fierce competitors in the rapidly developing frontier AI market. Rather than each company developing completely separate approaches to evaluating advanced AI systems, the proposed organisation could create common standards that would make it easier to assess increasingly capable models.
AI Safety Becomes a Bigger Global Issue
The reported discussions come as governments in the United States, European Union and Asia consider different regulatory frameworks for frontier artificial intelligence.
Governments and regulators have increasingly focused on how advanced AI systems should be tested before and after deployment, particularly as models become capable of handling more complex tasks across software development, research, business operations and other areas.
The renewed attention also follows public debate this month over the potential risks posed by increasingly powerful AI systems.
Anthropic co-founder and CEO Dario Amodei has argued that international cooperation will be necessary to address some of the most serious AI risks. His proposals have included international measures such as restrictions on the use of AI to develop biological weapons and shared verification systems designed to help countries monitor compliance.
Rival AI Labs Working Together
The reported agency would be notable because Google, OpenAI and Anthropic are competing directly for developers, enterprise customers, researchers and leadership in frontier AI.
A shared standards organisation could potentially give governments, businesses and researchers a common framework for understanding how advanced models perform and what safety testing they have undergone.
The development also reflects how quickly AI regulation is becoming an international issue. As countries develop their own rules, differences between national testing and reporting requirements could create additional challenges for companies operating globally.
For now, the reported Frontier AI Standards Agency remains an emerging initiative, but its formation would mark a significant development in the effort to establish common safety and testing standards for the next generation of artificial intelligence systems.