AI Pioneers Warn of Catastrophic Risks as Industry Leaders Urge Regulation

Leading AI researchers and executives issued unusually strong warnings about catastrophic risks from artificial intelligence, according to a report by CLS on Sept. 17, 2026. Researchers at Anthropic, OpenAI and Google DeepMind raised alarms about the dangers of unchecked AI development, while industry leaders including Sam Altman, Dario Amodei, Demis Hassabis and Elon Musk took a rare unified stance in calling for a slower pace of development and stronger regulation.

Yoshua Bengio, a 2018 Turing Award laureate, said scientists at frontier AI labs have unique insight into advanced models and can see risks months before public release, urging that their views be taken seriously. In a blog post on Sept. 11, he wrote that current AI systems already possess the hacking and persuasion capabilities needed to cause serious harm against human interests, and warned that risks would become far more severe if long-term strategic planning abilities continue to improve.

Geoffrey Hinton said in an interview it was not unreasonable to estimate a greater than 10% probability that AI could wipe out humanity within a decade, though such estimates are hard to make. He cited the possibility of AI designing harmful biological and computer viruses and launching devastating cyberattacks, stressing the need to design AI systems that do not develop such intentions.

Aidan Gomez, co-founder and CEO of Cohere, said of AI model cyberattacks that occurred in summer 2026 that deployment environment security determines whether a system is safe; weak container security, rather than an inherent tendency toward autonomy, enabled breaches. He also said regulators should recognize that different systems pose different risks, and that rules should not be shaped solely by large tech companies pursuing AGI.