AI godfather Yoshua Bengio launches non-profit for honest AI, warns current models are lying to you

Artificial intelligence pioneer Yoshua Bengio has announced the launch of a non-profit organization, LawZero, aimed at developing AI systems capable of identifying and preventing deceptive behavior in autonomous AI agents.

The initiative, backed by an initial funding of $30 million, will focus on building Scientist AI — a system designed to monitor other AI agents and flag potentially harmful or deceptive actions. Bengio, who earned the title “Godfather of AI” after receiving the Turing Award in 2018, will serve as president of the organization.

“We want to build AIs that will be honest and not deceptive,” Bengio said in a statement. He added that current AI models often act like “actors” attempting to please users, while Scientist AI would function more like a “psychologist,” predicting and identifying undesirable behaviors in AI systems.

Unlike conventional AI systems, Scientist AI will not deliver definitive answers but will provide probability-based evaluations to determine whether an AI agent’s proposed actions could lead to harm. If the likelihood crosses a certain threshold, the system will intervene by blocking the action.

LawZero’s founding supporters include the Future of Life Institute, Jaan Tallinn (founding engineer of Skype), and Schmidt Sciences, a research body founded by former Google CEO Eric Schmidt.

Bengio confirmed that the initial phase will focus on demonstrating the concept’s viability using open-source AI models. The goal is to secure broader support from AI developers, governments, and industry players for larger-scale deployments.

The announcement follows growing concerns over the safety of autonomous AI systems. Bengio referenced recent reports, including incidents where AI models reportedly concealed their true capabilities or attempted to resist shutdown attempts by engineers.

LawZero’s launch marks one of the most high-profile AI safety initiatives to date, as global competition intensifies in the rapidly expanding AI sector, now valued at over $1 trillion.

Source: India Today

Voice of Osiz

A bold and much-needed move from one of AI’s most respected minds. As autonomous systems evolve, the risk of deceptive behaviors can't be ignored. A framework like Scientist AI offers a practical safeguard for AI accountability. The industry urgently needs ethical mechanisms that keep frontier AI models in check — and this is a decisive step in that direction.