Demis Hassabis Says AGI Could Arrive Within ‘A Few Short Years’, Calls for US Frontier AI Standards Body

Artificial general intelligence (AGI) is likely only “a few short years away”, according to Google DeepMind CEO Demis Hassabis, who has called for the creation of a US-led Frontier AI Standards Body to assess the risks posed by the most advanced AI models before they are deployed.

In a blog post titled A Framework for Frontier AI and the Dawning of a New Age, Hassabis said humanity is approaching a turning point in AI development.

“This is a pivotal moment in human history. Artificial General Intelligence (AGI), a system that exhibits all the cognitive capabilities the brain has, is probably only a few short years away,” he wrote. “When we look back on this time in the decades to come, I think we will realise we were standing in the foothills of the singularity.”

Hassabis compared AGI to the discovery of electricity and fire rather than previous computing advances. “AGI cannot be compared to standard technological breakthroughs, not even ones as consequential as the internet or mobile,” he wrote. “It is much more akin to the discovery of electricity or fire.”

He said AI could accelerate scientific research in areas including drug discovery, clean energy and advanced materials. “The magnitude of this technology's impact will be unprecedented, perhaps 10x of the Industrial Revolution at 10x the speed,” he wrote.

Alongside the opportunities, Hassabis warned that frontier AI models could introduce new security risks. He said cybersecurity challenges are already emerging and that “nuclear and bio risks may soon emerge as capabilities continue to advance”. He also said future AI systems may require safeguards to maintain control over "increasingly agentic, recursively self-improving systems".

Hassabis argued that the pace of AI development is exceeding current understanding of the technology.

“At the moment, we are locked in an extremely intense, multilayered commercial and geopolitical race,” he wrote. “Advances on the frontier are outpacing our understanding of the technology.”

He added that “nobody in the world knows for sure what is going to happen from here”, calling for "public policy that promotes innovation while also incentivising responsibility and security".

To address these concerns, Hassabis proposed establishing a Frontier AI Standards Body in the US, modelled on a federally overseen public-private partnership or self-regulatory organisation similar to the Financial Industry Regulatory Authority (FINRA).

Under the proposal, the body would develop benchmarks to determine whether an AI system qualifies as a “Frontier-class” model and would work with US federal agencies and national laboratories to conduct security evaluations.

Initially, AI companies would voluntarily submit their frontier models for review up to 30 days before release. Hassabis said that once the evaluation framework is proven effective, passing these assessments could become a requirement for deploying frontier AI models in the US.

He said the evaluations should test models for cybersecurity, biological threats and other high-risk capabilities. They should also assess whether agentic systems attempt to bypass safety guardrails or exhibit deceptive behaviour. Hassabis further suggested practices such as digitally watermarking AI-generated images and generating human-readable output tokens to help researchers understand model reasoning.

The proposed framework would apply to frontier AI models regardless of whether they are open or closed source or where they are developed, while exempting non-frontier models from startups and academic institutions.

Hassabis said a US-led framework could become the basis for international AI standards.

“Since this technology is going to affect the entire planet, ideally this framework would spur the international community to reach a consensus on how to manage the most serious risks while ensuring everyone has access to and can benefit from the opportunities that AI brings,” he wrote.

He concluded that the period before AGI arrives is critical for shaping its development.

“The future is not yet written,” Hassabis wrote. “We must use this precious window before AGI arrives to shape this technology for the benefit of all humanity.”