AI Safety Concerns Highlighted at New Delhi Summit

Discussions at the AI Impact Summit in New Delhi in February 2026 focused on the risks associated with rapid artificial intelligence development, with experts urging increased oversight and government intervention.

Stuart Russell, a professor at the University of California, Berkeley, addressed the summit and described the lack of regulation in AI development as a procedural failure. He compared the current approach to AI safety to playing Russian roulette, emphasizing the potential global consequences if oversight is not strengthened.

Russell stated that competition among technology company leaders is driving an accelerated pace in AI research and deployment. He referred to this situation as an "arms race" and noted that some company executives may be unable to slow progress due to expectations from investors.

He reported that engineers at prominent AI laboratories have provided internal estimates suggesting the probability of human extinction from AI could be between 60% and 70%. These figures are higher than the 10% to 20% risk that has been publicly discussed by some in the field.

What the numbers show

• Some engineers at leading AI labs estimate a 60-70% risk of human extinction from AI
• Public risk claims for AI extinction are typically 10-20%
• Nuclear power safety standards require failure risk below one in 10 million per year
• AI industry risk levels remain unspecified according to Russell
• Hundreds of billions of dollars are being spent on AI data centers globally

Russell contrasted the safety requirements in the nuclear industry, where the annual risk of failure must remain below one in ten million, with the much less defined standards in the AI sector. He indicated that the current risk levels in AI development are not clearly quantified and are likely much higher than those in nuclear power.

He also stated that major investments are being made by both countries and private companies to build and operate energy-intensive data centers for AI training and deployment. This trend reflects the scale and speed of ongoing AI research and application worldwide.

Russell has over four decades of experience in artificial intelligence research and established a center at UC Berkeley in 2016 dedicated to AI safety and the development of systems that are demonstrably beneficial to humans.

The AI Impact Summit in New Delhi provided a platform for Russell to present his views and advocate for increased government involvement in regulating AI development. The event brought together stakeholders to discuss the challenges and responsibilities associated with advanced AI technologies.

* This article is based on publicly available information at the time of writing.