AI Standards Body OpenAI Google Anthropic: Can the Industry Regulate Itself?

OpenAI, Anthropic and Google are working toward the creation of an AI standards body, according to recent reports.

The idea originated from Google DeepMind CEO Demis Hassabis, who proposed an organization that could test powerful AI systems before they are released.

The concept resembles self-regulatory organizations used in other industries.

The proposed body would focus on technical testing and safety rather than becoming a traditional government regulator.

Why is this happening now?

AI systems are becoming increasingly capable.

They can write software, operate tools, interact with external systems and perform increasingly complex tasks.

That makes independent testing more important.

A model developer obviously has an incentive to demonstrate that its system performs well.

An external standards organization could provide another layer of evaluation.

OpenAI has also recently called for international coordination around AI safety standards and testing.

But self-regulation comes with obvious questions.

Who controls the organization?

Who pays for it?

What standards does it use?

Can companies be trusted to regulate competitors?

And how does an industry-led body interact with government regulators?

Those questions don't make the idea unworkable.

They simply show that technical standards and regulation are different things.

A standards body could establish common testing methods.

Governments could still decide what is legally permitted.

For businesses building AI systems, this could eventually lead to something useful: common evaluation standards.

Instead of every company inventing its own definition of AI safety, there could be shared benchmarks for capability, security and risk.

The idea is still developing.

But the discussion itself shows something important:

AI companies are beginning to recognize that increasingly capable models need shared rules for testing.