Published December 5, 2025 · Updated December 5, 2025
Demis Hassabis, CEO and co-founder of Google DeepMind, warned during a recent conference that artificial general intelligence (AGI) could become “transformative” as early as 2030. Speaking to Axios, Hassabis stressed that the world must begin coordinated discussions on safety, governance and long-term risk — before AGI capabilities accelerate beyond existing guardrails.
The remarks highlight a shift among leading AI researchers: AGI is no longer an abstract future concept, but an emerging technology with real-world implications for economies, security, ethics and global stability.
Key Takeaways
- DeepMind CEO Demis Hassabis warns AGI could become “transformative” by 2030.
- Hassabis calls for urgent global dialogue around AGI safety, governance and international cooperation.
- AGI development is accelerating faster than anticipated due to multimodal models, breakthroughs in scaling and improved compute.
- Experts emphasise building shared standards before capabilities exceed existing oversight.
- Growing pressure on governments to align safety frameworks ahead of AGI-level systems.
Explore More
Looking to go deeper? Explore Arti-Trends’ core knowledge hubs for technical insights, practical tools, real-world applications and strategic AI analysis:
- AI Guides Hub — foundational explanations and deep technical breakdowns
- AI Tools Hub — hands-on comparisons and evaluations
- AI News Hub — rapid updates on global AI developments
- AI Investing Hub — market-focused insights into AI companies and next-gen infrastructure
These hubs give you broader context behind the models, theories and safety frameworks shaping the AI landscape.
What Hassabis said about AGI
During a conversation with Axios, Hassabis noted that recent breakthroughs in multimodal LLMs, reinforcement learning and large-scale compute “have dramatically accelerated the timeline” toward systems with general capabilities.
According to him, AGI is likely to:
- reason across domains
- combine perception, language and planning
- execute complex tasks autonomously
- outperform experts in scientific and creative domains
- unlock new classes of research and technological innovation
Hassabis stressed that while AGI could solve major global challenges — climate modelling, drug discovery, logistics, education — the stakes are extremely high if safety is not prioritised early.
“We need a global conversation about how to build this responsibly,” Hassabis said, suggesting formal coordination between governments, labs and independent safety experts.
Strategic context & industry impact
The AGI debate is intensifying as top labs — DeepMind, OpenAI, Anthropic, xAI — push rapidly toward increasingly general systems.
The implications are wide-ranging:
For regulators & governments
- Need for international AGI-safety frameworks
- Coordination similar to nuclear or biotech governance
- Risk of geopolitical competition driving unsafe acceleration
For companies & enterprise users
- AGI-powered automation may radically shift productivity
- Strategic planning must account for new capabilities (and risks)
- Safety-certified AI models may become regulatory requirements
For researchers & the AI community
- Pressure to slow deployment until safety improves
- Growing divide between open-source vs closed AGI pathways
- Critical need for alignment research, interpretability and red-teaming
Why AGI may be closer than expected
Hassabis pointed to four accelerating trends:
- Massive advances in multimodal reasoning
Models now combine vision, language, audio, coding, planning and robotics. - Scaling laws still holding
More data + more compute + better architectures = predictable capability jumps. - Frontier training runs increasing
Companies prepare trillion-parameter-scale and multi-agent-scale systems. - Scientific capabilities emerging
Models show promise in protein design, materials science and physics research — key AGI benchmarks.
These trends support the idea that AGI is no longer speculative but a realistic near-term milestone.
Safety concerns & governance debates
Hassabis’ call echoes concerns from policymakers and safety researchers:
- AGI could amplify cyber risks, misinformation and autonomy failures
- Misaligned agents may behave unpredictably
- Unequal access to AGI could widen global inequality
- Powerful AGI may centralise control within a handful of companies
The central question:
How do we ensure AGI benefits humanity without creating systemic risk?
Hassabis argues the answer lies in:
- pre-AGI governance agreements
- coordinated testing and red-team frameworks
- transparency around model capabilities
- cultural and international alignment
- multi-stakeholder oversight, not corporate control alone
What happens next
DeepMind is expected to publish more formal guidance on AGI safety later this year.
Governments in the US, EU and UK are watching closely as frontier AI labs move toward increasingly general systems.
If Hassabis’ 2030 timeline holds, AGI could soon become the most powerful — and most debated — technology in human history.
For more analysis on AGI models, safety frameworks and global AI governance, explore related guides in the AI Guides Hub and follow rapid updates in the AI News Hub.
Source
Axios — DeepMind CEO: AGI could be transformative by 2030, urges global safety dialogue
(December 2025)
