NextFin News - In a significant intervention at the intersection of technology and global policy, Demis Hassabis, the Chief Executive Officer of Google DeepMind, has called for an immediate and urgent increase in research dedicated to the existential and operational threats posed by artificial intelligence. Speaking at the India AI Impact Summit 2026 in New Delhi, which concluded on Saturday, February 21, Hassabis emphasized that the industry is at a critical juncture where the speed of innovation is outstripping the capacity of global regulators to ensure safety.
According to Bernama, Hassabis identified two primary categories of risk that require immediate academic and industrial focus: the exploitation of AI by malicious actors and the potential for humans to lose control over increasingly autonomous systems. During an exclusive interview at the summit, Hassabis argued for the implementation of "smart regulation" designed to mitigate real-world risks without stifling the beneficial potential of the technology. He was joined in this sentiment by OpenAI CEO Sam Altman, who also urged for swift regulatory frameworks to keep pace with the rapid evolution of generative and autonomous models.
The summit, hosted by India, served as a platform for Prime Minister Narendra Modi to advocate for international cooperation to ensure AI serves as a tool for human benefit. However, the call for a unified global approach met significant resistance from the United States. Michael Kratsios, leading the U.S. delegation, reiterated that the administration of U.S. President Trump remains firmly opposed to any form of global AI governance, prioritizing national sovereignty and competitive dominance over international regulatory alignment. This friction underscores a growing divide between the tech industry's leadership, which is increasingly wary of the monsters it may be creating, and a U.S. executive branch focused on a 'light-touch' domestic approach to maximize American technological leadership.
The urgency expressed by Hassabis is rooted in the accelerating capabilities of Large Language Models (LLMs) and autonomous agents that have emerged since the 2024-2025 boom. As of early 2026, AI systems are no longer merely generating text or images; they are increasingly integrated into critical infrastructure, financial markets, and cybersecurity defense. The "loss of control" scenario Hassabis warned about refers to the 'alignment problem'—the technical challenge of ensuring that an AI's goals remain perfectly synchronized with human intent as the system's reasoning becomes more complex and opaque.
From a financial and industry perspective, the call for more threat research is a strategic move to preempt catastrophic failures that could lead to a total collapse of public trust or draconian, reactive legislation. By advocating for "smart regulation" now, leaders like Hassabis and Altman are attempting to shape the regulatory landscape in a way that favors established players who have the resources to implement complex safety protocols. This 'regulatory capture' through safety advocacy is a common framework in high-stakes industries, yet the technical reality of AI threats makes the concern genuine. Data from the 2025 AI Safety Index suggests that while corporate investment in AI capabilities grew by 40% year-over-year, investment in safety and alignment research lagged at less than 5% of total R&D spend.
The stance of the U.S. government under U.S. President Trump represents a significant hurdle for the "global governance" camp. By rejecting international oversight, the U.S. is betting that a decentralized, market-driven approach will allow American firms to innovate faster than their counterparts in more regulated jurisdictions like the European Union. However, this creates a 'race to the bottom' risk where safety standards are sacrificed for speed. If a major AI-driven incident occurs—such as a systemic financial flash crash or a large-scale automated cyberattack—the lack of a coordinated international response framework could exacerbate the fallout.
Looking ahead, the tension between the tech sector's call for safeguards and the U.S. administration's deregulatory stance is likely to intensify. We can expect a shift where AI safety research becomes a competitive moat; companies that can prove their systems are "provably safe" will win lucrative government and enterprise contracts. Furthermore, as India and other nations in the Global South continue to host summits like the one in Delhi, we may see the emergence of a non-Western regulatory bloc that establishes its own standards, potentially fragmenting the global AI market into different "safety zones."
Ultimately, the warnings from Hassabis suggest that the industry is no longer in its 'move fast and break things' phase. The stakes have shifted from breaking software to potentially breaking societal structures. As U.S. President Trump continues to push for American AI supremacy, the burden of safety may fall increasingly on the private sector and academic researchers to self-police, a prospect that Hassabis clearly views as insufficient without the backing of robust, albeit smart, governmental oversight.
Explore more exclusive insights at nextfin.ai.
