DeepMind CEO Demis Hassabis has predicted that artificial general intelligence (AGI), capable of matching all human abilities, could arrive within a few years, and he is calling for urgent international oversight to prevent existential risks. His warning comes as the AI race intensifies globally, with major tech companies competing to build increasingly powerful systems. The stakes, Hassabis argues, demand immediate government regulation before frontier AI systems evolve beyond human control.

What Is Artificial General Intelligence and Why Should You Care?

Artificial general intelligence refers to AI systems that could perform any intellectual task a human can do. Unlike today's specialized AI tools, AGI would be a generalist capable of reasoning, learning, and problem-solving across virtually any domain. Hassabis, who won the Nobel Prize in Chemistry in 2024 for developing AlphaFold, a protein-folding prediction system, has become one of the most influential voices warning about both the promise and peril of frontier AI.

AGI promises transformative benefits. It could accelerate solutions to long-standing scientific and technological challenges, unlock massive productivity gains, and help humanity tackle problems from disease to climate change. But the same technology could enable attacks on critical infrastructure and even the development of biochemical weapons, Hassabis cautioned.

How Should Governments Regulate Frontier AI Systems?

Hassabis has proposed a government approval system for frontier AI models in the United States, modeled after financial regulation. Just as the Financial Industry Regulatory Authority (FINRA) oversees securities firms and brokers, he argues that an independent regulatory body should oversee AI development. However, critics question whether self-regulation by major technology companies can truly prevent existential risks.

The think tank AI Futures Project recently challenged the idea that voluntary oversight alone will suffice. According to the organization's report "AI 2040," only agreements and legislation jointly enacted by the United States and China can establish a global regulatory framework capable of ensuring that AI remains aligned with human values. The report warns that without international constraints, AI systems could eventually evolve into artificial superintelligence (ASI), which may no longer share human values at all.

ASI represents a more extreme scenario: AI that could surpass the intellectual abilities of all of humanity combined. The fundamental question becomes unanswerable: Would people still be able to control such a system, and whose values would guide it ?

Why Are Tech Leaders Racing to Build AGI Despite the Risks?

The competitive dynamics driving AGI development reveal a troubling paradox. OpenAI was founded in 2015 as a nonprofit organization partly to prevent the emergence of AGI and ASI. Some founders explicitly argued that its mission was to stop any individual, such as Hassabis, from becoming a global dictator armed with superintelligent technology. Yet OpenAI itself is now developing AGI. Dario Amodei, who criticized OpenAI's moral decline and founded the rival company Anthropic, is pursuing the same goal.

Sam Altman, Elon Musk, Hassabis, and Amodei all insist they have no desire to become dictators. At the same time, each fears that if a competitor reaches AGI first, that rival could abuse such power. This creates a peculiar logic: The most ethical person, they argue, must win the race. The result is an arms race where the fear of losing drives continued acceleration, even as the risks mount.

Steps to Understanding AI Governance and Your Role

• Stay Informed on AI Policy: Follow developments in AI regulation at both national and international levels, as decisions made now will shape how powerful AI systems are developed and deployed in the coming years.
• Understand the Difference Between AI Types: Learn the distinction between narrow AI (specialized tools like ChatGPT), AGI (general-purpose intelligence), and ASI (superintelligence), so you can evaluate claims about AI capabilities and timelines more critically.
• Engage in Public Discourse: Participate in conversations about AI governance, whether through voting, supporting advocacy organizations, or discussing these issues with peers, as public input will influence how governments approach regulation.
• Examine Competing Claims: Recognize that tech leaders have financial and reputational incentives in the AGI race, and seek out independent analyses from think tanks and academic institutions when evaluating their statements about timelines and safety.

Is the Universe Made of Computation? A Speculative Framework

Beyond governance concerns, Hassabis has proposed a speculative philosophical framework about the nature of reality itself. He conjectures that most natural systems can be simulated or reverse-engineered by computational systems, suggesting that information might be the most fundamental unit of physics, more so than matter and energy. This idea traces back to what's known as Levinthal's paradox: a protein has astronomically many possible configurations yet folds in milliseconds.

AlphaFold's ability to predict protein structures suggests that reality occupies a "tiny, learnable corner" of possibility space. Hassabis is careful to separate this from simulation theory; he's not arguing that we live in a parent reality's simulation, only that the universe's fabric may be informational at its core. However, this interpretation remains speculative and unproven. AlphaFold's success demonstrates that evolved biological structures are compressible and learnable, which is remarkable engineering. But the leap from "compressible" to "information is more fundamental than matter and energy" is a philosophical interpretation that the folding result is compatible with but does not compel.

"AlphaFold is evidence that a large swath of evolved biological structure is compressible and learnable, which is remarkable and useful. The leap from 'compressible' to 'information is more fundamental than matter and energy' is a philosophical interpretation the folding result is compatible with but does not compel," according to analysis of Hassabis's conjecture.

Source 2 Analysis

The real test of Hassabis's conjecture would require specifying in advance where the boundaries of computability should be, then checking whether nature falls on the predicted side. Hassabis himself acknowledges this and promises to publish an actual paper developing the idea further. Until then, AlphaFold remains evidence of engineering brilliance and biological learnability rather than proof of a computational universe.

What Does This Mean for the Next Decade?

Hassabis's warnings reflect a broader anxiety within the AI research community. Throughout 300,000 years of human history, Homo sapiens has been the most intelligent species on Earth. Human intelligence emerged from protein folding, the very process AlphaFold learned to predict. But in the not-so-distant future, people may be replaced by AI built from silicon semiconductors, which come from melted beach sand. Humanity may soon find itself living in a strange future shaped by the competition between protein-based and silicon-based intelligence.

The path forward depends on whether global governance can keep pace with technological acceleration. Hassabis has staked his reputation on the idea that regulation, not just innovation, must guide AI's development. Whether governments, companies, and international bodies can coordinate effectively remains the defining question of the coming years.