AGI in two to three years: DeepMind's Hassabis shortens his forecast again
Representative image: axonite / Pixabay
Google DeepMind chief Demis Hassabis now expects AGI in two to three years, down from three to five, according to a profile published on 21 October 2026, while warning safety is lagging.
What Happened
AGI, or artificial general intelligence, could arrive within two to three years, according to Demis Hassabis, the chief executive of Google DeepMind and a Nobel laureate in chemistry. His new estimate was reported in a Wall Street Journal Magazine profile published on 21 October 2026, in which he said he was "updating nearer and nearer". Earlier in 2026, Hassabis had put AGI three to five years away, and before that further out. The shorter forecast comes days after he warned that safety and security measures for advanced AI are not keeping pace, describing recent incidents in which AI agents were used for hacking as "early warning shots".
Key Facts
- New estimate: AGI in two to three years, which would mean roughly 2028 or 2029, as reported on 21 October 2026.
- Previous estimate earlier in 2026: three to five years, according to reports comparing his statements.
- Reports of his earlier remarks this year vary: five to ten years at Davos in January, five to eight years in New Delhi in February, and 2030 at a Sequoia conference in April.
- Hassabis defines AGI as a system that can match the full range of human cognitive abilities across many different tasks.
- On safety, he said at a summit in Dublin that AI-agent hacking incidents are "early warning shots" of a possible security crisis, and wrote on X on 20 October that "the safety and security challenges in front of us as we confront superintelligence are immense".
- The estimate is reported in the magazine's words rather than as a full quotation, and different labs define AGI differently.
Why It Matters
Hassabis is one of the most closely watched voices in AI. He co-founded DeepMind, which built AlphaGo and AlphaFold, the system that predicts the shapes of proteins and earned him a share of the 2024 Nobel Prize in Chemistry. He has also been seen as more cautious than some rivals about timelines. When someone with that record cuts his estimate twice in a year, it signals how quickly leading labs believe their systems are improving.
The term AGI needs care. There is no single agreed definition. Some companies use it to mean AI that can do most office work as well as a person; Hassabis uses a stricter test of matching the full breadth of human thinking, from reasoning to creativity. A shorter forecast under a strict definition is therefore more striking than a similar claim using a looser one. Critics point out that forecasts in this field have often been wrong in both directions.
The safety warning is the other half of the story. Hassabis has argued that the industry and governments need to move faster on security standards, especially for AI agents, programs that can take actions on computers and the internet on a user's behalf. If more capable systems arrive sooner, the time to put rules, testing and safeguards in place becomes shorter.
For India, which has positioned itself as a major user and builder of AI through the IndiaAI Mission and which hosted Hassabis at an AI summit in New Delhi in February, a faster timeline raises practical questions: how quickly jobs in IT services and back-office work may change, and how ready regulators are to test powerful models before they are widely used.
| When | Hassabis's AGI estimate (as reported) |
|---|---|
| January 2026, Davos | five to ten years |
| February 2026, New Delhi | five to eight years |
| April 2026, Sequoia conference | 2030 |
| October 2026, magazine profile | two to three years |
Impact
Short-term: The new estimate will feed debates among investors, companies and policy-makers about how fast to spend on AI and how quickly to regulate it.
Long-term: If AGI as Hassabis defines it arrives near the end of this decade, it would affect almost every sector, from science and medicine to education and employment, and would make international agreements on AI safety more urgent.
Who is affected: Technology companies and their workers, governments writing AI rules, researchers, students choosing careers, and the IT services industry in India.
Key Takeaway
Google DeepMind's Demis Hassabis now says AGI could arrive within two to three years, his second cut this year, while warning that AI safety and security work is falling behind.
Questions and Answers
When does Demis Hassabis think AGI will arrive?
In two to three years, according to a profile published on 21 October 2026. That would put it around 2028 or 2029, down from his earlier 2026 estimate of three to five years.
What is AGI?
Artificial general intelligence is AI that can match human ability across a wide range of thinking tasks rather than one narrow job. Hassabis defines it as a system with the full range of human cognitive abilities; other companies use looser definitions.
What did Hassabis say about AI safety?
He described AI-agent hacking incidents as "early warning shots" of a possible security crisis and wrote on 20 October 2026 that the safety and security challenges of superintelligence are immense.
Is the two-to-three-year AGI forecast certain?
No. It is one expert's estimate, reported in a magazine's words, and AI forecasts have often been wrong. Labs also disagree on what counts as AGI.
Disclaimer: Prepared by the Peepals newsroom from publicly available sources with AI assistance. Information is accurate to the best of our knowledge at the time of publication and may change. Images may be representative. Not professional advice. Report an error via our contact page.
Sourced and fact-checked by the Peepals Global Editorial Team
Vultr and AMD plan 20 MW AI supercluster in Chennai with MI355X chips
Goa AI Policy 2026 aims to train 50,000 people in AI by 2030
Reported, fact-checked and published by the Peepals Global Editorial Team.
