AGI could emerge by 2030, bringing world-changing potential and existential risks. DeepMind urges global oversight for safe development.
Artificial General Intelligence (AGI)—a form of AI with human-like cognitive abilities—could be developed by as early as 2030, and it might bring with it not only world-changing advancements but also the potential to “permanently destroy humanity,” according to a new research paper published by Google DeepMind.
The stark warning comes from a comprehensive study co-authored by DeepMind co-founder Shane Legg and supported by CEO Demis Hassabis, outlining the risks associated with AGI and advocating for global governance to ensure safe development.
“Given the massive potential impact of AGI, we expect that it too could pose potential risk of severe harm,” the DeepMind study warns, noting that “existential risks that permanently destroy humanity are clear examples of severe harm.”
What Is AGI and Why It Matters
AGI, unlike traditional narrow AI models (such as language translation tools or recommendation systems), aims to mimic the versatility of human intelligence. It can perform a wide range of intellectual tasks, adapt to new environments, and even improve itself autonomously.
In short, AGI represents the transition from machines that assist to machines that can think—a shift that could be revolutionary or catastrophic, depending on how it’s managed.
DeepMind’s Categorization of AGI Risks
The DeepMind paper doesn’t elaborate on the exact ways AGI could destroy humanity but instead categorizes the risks into four major domains:
- Misuse– where AGI is intentionally used to harm people.
- Misalignment– when AGI’s goals do not align with human values.
- Mistakes– unintended outcomes due to system errors.
- Structural Risks– broader societal disruptions and systemic collapse.
While some may dismiss these risks as speculative, the paper strongly emphasizes that they are real and must be addressed early.
“In between these ends of the spectrum, the question of whether a given harm is severe isn’t a matter for Google DeepMind to decide,” the paper notes, “instead it is the purview of society, guided by its collective risk tolerance and conceptualization of harm.”
Demis Hassabis: Call for a UN-Like Oversight Body
In February, Demis Hassabis, CEO of DeepMind, reiterated the urgency for collaborative international oversight. He proposed a framework that mirrors existing global institutions in science and security.
“I would advocate for a kind of CERN for AGI,” Hassabis said, “and by that, I mean a kind of international research-focused high-end collaboration on the frontiers of AGI development to try and make that as safe as possible.”
Hassabis further elaborated on the need for a dual system: one focused on research and the other on regulatory enforcement.
“You would also have to pair it with a kind of institute like IAEA, to monitor unsafe projects and sort of deal with those,” he said. “And finally, some kind of supervening body that involves many countries around the world that input how you want to use and deploy these systems. So kind of like a UN umbrella—something fit for purpose, a technical UN.”
Shane Legg’s Long-Held Warnings
Shane Legg, who co-authored the DeepMind paper and serves as the company’s Chief AGI Scientist, has consistently warned about the rapid pace of AGI development. As early as 2011, Legg predicted there was a 50% chance AGI could be realized by 2028—a prediction he still stands by.
In an interview with Time, Legg said, “Many experts who previously dismissed such timelines are now revising their views, and with good reason. We need to start preparing now.”
Although the DeepMind paper refrains from detailing how AGI might lead to humanity’s extinction, Legg’s perspective supports the urgency to set up robust safeguards well in advance.
Why the Timeline Matters
Estimates on when AGI will arrive vary, but most now agree it’s not a matter of if, but when. Hassabis, Musk, and other leaders believe it could emerge within a decade.
“AGI, which is as smart or smarter than humans, will start to emerge in the next five or 10 years,” said Hassabis.
Meanwhile, massive investments by tech giants like OpenAI and Amazon show that the AI race is accelerating. OpenAI recently raised $40 billion, while Amazon’s AGI Lab introduced Nova Act, an AI agent that outperforms competitors in complex reasoning and web interactions.
Urgency for Ethical and Regulatory Frameworks
As AI systems inch closer to human-level intelligence, ethical questions take center stage. Who decides how AGI is used? What rights do AI systems have, if any? What happens if it acts against human interest?
Experts argue that the world must not repeat the mistakes of past technological revolutions by failing to plan for unintended consequences.
The DeepMind study encourages governments, civil society, and AI developers to come together and establish shared protocols for transparency, accountability, and global safety.
Final Thoughts
The rise of AGI could represent humanity’s greatest invention—or its greatest threat. As Google DeepMind warns, ignoring the risks could be catastrophic.
“We must proceed with both ambition and caution,” said Hassabis. “The path to AGI is one of enormous potential, but also enormous responsibility.”
If AGI does arrive by 2030, how we prepare today may determine whether it becomes a tool for collective progress—or a force that leads to irreversible loss.
Read more: Wipro Appoints Chandna Raja G as Director – Head of Talent Acquisition
