Legg Opens DeepMind Institute, Calls AGI Claims Premature

Shane Legg, the DeepMind co-founder who has said for years that an AGI-like system is roughly a coin flip away by 2028, used the launch of a new institute to tell the Financial Times that the field's loudest recent claim, that AGI is already here, is wrong.

According to the FT report as syndicated by Reuters, Legg called it "premature to declare that AGI had been achieved, despite recent claims by executives at Nvidia and OpenAI." He warned that "rapidly advancing AI must not outpace safety controls." The quotes come to the wider press through the FT rather than as direct speech, so read them as reported.

The occasion was the DeepMind Institute, which Legg is opening to broaden public discussion of AGI, which Google DeepMind defines as "a system with the cognitive capabilities of the human brain." Its scope covers "science, education, society, policy, philosophy and human flourishing." Its first three essays address AI safety, economic policy and principles for a "new utopianism."

Legg's own forecast has not moved: he remains "comfortable" with a 50% probability of "minimal" AGI by 2028. What has shifted is his willingness to talk about pacing in public. Anthropic CEO Dario Amodei's proposal to slow but not pause the release of frontier models was, per the FT paraphrase, "interesting directionally" and "worth considering." That endorsement lands the same week our tracker logged OpenAI and Anthropic staff being blindsided by a 'Pace the Frontier' plan and ControlAI's hundreds of Hill meetings on extinction risk.

The underlying worry Legg names is agentic systems, ones that plan and execute tasks with less human hand-holding, advancing faster than the tools used to test and monitor them.

The FT report and its syndications do not spell out the institute's funding structure, its staff, or how independent it is of DeepMind's commercial roadmap. Legg's remarks reach the record as paraphrase rather than direct quotation.