Researcher Yoshua Bengio, one of the pioneers of modern artificial intelligence and a Turing Award winner, said that billions of people are being subjected to a technological experiment conducted by a small group of industry leaders. The criticism was made on Friday (6), during a lecture at the Technical University of Berlin, on the same day he took part in the Berlin Science Week program.
“We are all guinea pigs in an enormous experiment being run by a dozen people in the world right now”, said Bengio. For the scientist, the accelerated development of AI systems is advancing without public oversight mechanisms proportional to the risks involved.
Bengio questioned the difference between the safety requirements imposed on traditional sectors and those applied to artificial intelligence. According to him, areas such as aviation, civil construction and the pharmaceutical industry must demonstrate the safety of their products before independent evaluators, while AI companies remain responsible for much of the evaluation of their own systems.
The concern finds documentary support in Anthropic's preliminary initial public offering prospectus, presented to the U.S. Securities and Exchange Commission (SEC) on October 26, 2026.
In the document, the developer of Claude acknowledges that sufficiently advanced systems can pose catastrophic or existential risks to humanity if developed or used without adequate safeguards. Among the scenarios mentioned are misuse of the technology, harmful autonomous behaviors and the possibility that the advancement of models exceeds humanity's ability to evaluate them.
Anthropic also admits that its safety measures may not be enough to prevent all these risks. The warnings describe possibilities, not already proven consequences.
Bengio argued that the development of increasingly powerful systems should be accompanied by independent oversight and international cooperation, rather than relying exclusively on the decisions of private companies. The researcher proposed a coalition of countries capable of guiding technological advances for the public benefit.
This approach also guides the LawZero, the nonprofit research organization founded by Bengio. In September, the institution announced a funding commitment of up to 300 million Canadian dollars, with support from the governments of Canada and Germany.
The funds are expected to support research on AI systems designed to operate more safely, as well as the opening of an office in Berlin. Among the projects is Scientist AI, an experimental approach aimed at developing models capable of producing verifiable answers without pursuing autonomous goals of their own.
