Standing before a government audience in New Delhi in June 2026, Demis Hassabis delivered a sentence that compressed the future into a handful of years. “In 2026 AGI is on the horizon, maybe within the next five years,” the Google DeepMind CEO said during his keynote at the India AI Impact Summit, an event organized by the Indian government. The remark, captured in an official press release from India’s Press Information Bureau, places 2029 at the near edge of his window and puts one of the world’s most consequential AI researchers on record: the technology his lab is building could reach human-level general intelligence before the end of this decade.
That timeline is not an outlier among frontier lab leaders, but it is notable for who is saying it and where. Hassabis runs the lab behind AlphaFold, AlphaGo, and the Gemini family of models. He is not a commentator speculating from the sidelines; he is directing the research. And he chose to deliver the message at a government summit, not a tech conference, signaling that his audience is now policymakers as much as engineers.
A pattern of accelerating forecasts
The New Delhi remarks fit a trajectory Hassabis has been building in public for at least a year. In an interview with The Guardian published in August 2025, he argued that the coming wave of AI would be larger than the Industrial Revolution and arrive roughly ten times faster. That was not a throwaway comparison. He used it to make a structural point: universities, regulatory agencies, and labor markets are all built for a pace of change that AI has already outstripped. The India summit statement sharpens the same argument by attaching a number to it.
Hassabis is not alone in pulling the timeline forward. Anthropic CEO Dario Amodei has spoken publicly about powerful AI systems arriving by 2026 or 2027. OpenAI’s Sam Altman has suggested AGI could emerge within a few years. But the field is far from unanimous. Meta’s chief AI scientist, Yann LeCun, has repeatedly argued that current architectures lack the reasoning and world-modeling capabilities true AGI would require, and that the milestone could be decades away. The gap between these positions is not just a scheduling disagreement; it reflects deep divisions over what AGI actually means.
The definition problem
No full transcript or video of Hassabis’ India keynote has been released beyond the single paragraph in the PIB document. That means the broader context of his remarks, including any caveats or technical definitions he may have offered, is unavailable for independent review. The omission matters because the term “artificial general intelligence” carries wildly different meanings depending on who uses it.
For some researchers, AGI means a system that can pass any professional licensing exam or standardized test a human can. For others, it means a system capable of autonomously designing and running novel scientific experiments. The first definition is arguably within reach of today’s large language models on certain benchmarks; the second would represent a qualitative leap that no existing system has demonstrated. Which version Hassabis had in mind shapes how seriously the five-year window should be taken, and without more detail, outside observers are left weighing his track record against the inherent difficulty of predicting discontinuous breakthroughs.
Governments are moving, but slowly
While Hassabis was speaking in New Delhi, governments on multiple continents were scrambling to build regulatory scaffolding. Earlier in 2026, the White House released a national AI legislative framework under President Donald J. Trump, laying out governance priorities for AI development in the United States. The document signals that the federal government recognizes the need for structured legislation, but the publicly released summary does not specify enforcement mechanisms, funding levels, or compliance deadlines. It establishes intent without yet creating binding obligations for AI developers.
The framework does not reference AGI timelines or Hassabis by name, and treating it as evidence that Washington agrees AGI is imminent would be a misreading. It addresses AI broadly. Still, its existence confirms that the policy conversation has shifted from “should we regulate?” to “how fast can we build the rules?” If Hassabis’ timeline is even roughly correct, the window for converting high-level frameworks into enforceable law is measured in years, not decades.
India’s decision to host the summit itself reflects a similar urgency. The country is positioning itself as both a developer and a governance voice in AI, and inviting the head of DeepMind to keynote a government event suggests New Delhi wants a seat at the table before the rules are written elsewhere.
What the evidence supports, and what it does not
The strongest piece of evidence here is the PIB release: a primary government document recording a named speaker’s words at an official event. When Hassabis says AGI could arrive within five years, readers can be confident he actually said it in a formal, recorded setting. The Guardian interview, published months earlier, corroborates that he has been building this argument publicly over time rather than floating a one-off prediction.
What the evidence does not support is any precise forecast about what AGI will look like when it arrives, how it will be measured, or how societies will respond. Hassabis may be extrapolating from internal DeepMind research milestones that remain confidential. AI capability benchmarks have improved sharply in recent years, but the relationship between benchmark performance and genuine general intelligence is itself a matter of active scientific debate.
For policymakers, the prudent reading is that timelines may be shorter than many existing strategies assume. Building regulatory capacity, retooling education systems, and stress-testing economic and national security institutions are multi-year projects. If AGI-level systems emerge closer to 2029 than 2039, the gap between starting that work now and waiting for certainty could prove decisive.
The race between the technology and the rules
Hassabis’ comments ultimately force a question that no government has fully answered: what happens when a technology that could reshape every sector of the economy arrives faster than the institutions meant to manage it can adapt? The technology curve is steepening. National frameworks are being drafted. But between a five-year AGI window and a policy apparatus still debating definitions and jurisdictions, the distance is vast. Whether that gap closes in time is not a technical question. It is a political one, and the clock Hassabis described in New Delhi is already running.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
