DeepMind co-founder warns that AI capabilities should not outpace safety controls.

What it means

Shane Legg, a co-founder of Google DeepMind, warned in an interview with the Financial Times that the rapid advancement of AI capabilities must never outpace the development of safety controls. He made the remarks while launching a new 'DeepMind Institute' intended to foster debate and research on artificial general intelligence (AGI) safety, responding to growing industry concerns about recursive self-improvement in AI systems.

Accuracy & sourcesGrounded

The claim is solid: it is a direct warning attributed to a named co-founder of DeepMind, reported by the Financial Times, a reputable secondary source. The reporting sits close to the primary source—Legg's own statements during the launch of the institute—so the core assertion is well-supported.

How new

Genuinely new: the Financial Times article was published on 16 September 2026, and this warning is part of a fresh announcement around the DeepMind Institute launch.

Go deeper