Google’s James Manyika calls for shared responsibility in AI regulation

One of Google’s top executives says AI is risky, needs regulation, and that the job of regulating it can’t fall on any single group. That’s a notable message from a company that builds some of the most powerful AI models on the market.

James Manyika, Google’s senior vice president overseeing research and technology, made the case in an interview that aired on October 2, 2026. His argument was simple: everyone who touches AI owns a piece of the safety problem.

“AI’s risks are real, regulation is necessary and responsibility has to be shared across industry, government and society.”

Who owns the safety problem

Manyika drew a clear line on accountability. Companies building frontier AI models, the most advanced and capable systems, should be responsible for making them safe.

But he didn’t stop at the labs. He argued that deployers, users and policymakers all have roles to play too.

Manyika also pushed back on the idea that industry should wait for government to set the rules first. He said waiting for President Trump to regulate AI made little sense, and argued that voluntary industry efforts should run in parallel with whatever governments eventually put in place.

The voluntary accord and its limits

Manyika’s comments land as the AI industry leans further into self-regulation. Major tech firms, including Google, OpenAI and Nvidia, recently signed a voluntary accord aimed at strengthening safety practices.

The agreement centers on what participants describe as “robust internal processes” alongside independent auditing.

The catch: the accord carries no legal enforcement mechanisms. If a company falls short of its own commitments, there’s no regulator waiting with a fine.

A longer-running argument

Throughout 2025 and 2026, Manyika has argued that AI is too important to leave unregulated, and equally too important to regulate badly.

His preferred approach involves balanced, enabling frameworks. Those are rules designed to guard against real harms without choking off useful progress.

He has pointed to sectors like health and infrastructure as areas where AI could deliver major benefits.

Manyika also co-directs the newly established DeepMind Institute, giving him a foot in both the research and policy sides of Google’s AI operation.

What this means

For the AI industry, Manyika’s comments signal that leading companies want to be seen as proactive on safety rather than reactive.

Critics of self-regulation will likely focus on the enforcement gap. A voluntary accord with independent audits is a step, but without legal teeth, accountability ultimately depends on corporate goodwill and reputational pressure.

The thing to watch is whether the voluntary approach eventually hardens into enforceable standards, and who gets to write them. If companies define the baseline first, governments may find themselves codifying rules the industry already drafted.