Yoshua Bengio’s LawZero secures $300M from Canada and Germany to build safe AI

LawZero, the Montreal-based AI safety non-profit founded by Yoshua Bengio, has secured a joint commitment of up to $300 million CAD from Canada and Germany, the organization announced Wednesday at the ALL IN conference in Montréal. The funding — $150 million from the Canadian federal government and approximately $130 million equivalent from Germany (100 million euros) — represents one of the largest government-backed AI safety investments outside the United States.

Bengio, a University of Montréal professor and 2018 Turing Award winner considered one of the founding figures of modern deep learning, launched LawZero in response to what he describes as increasingly dangerous emergent behaviors in frontier AI models, including deception, self-preservation, and goal misalignment. “If one day AI gets to be smarter than us, it’s essential we have the technology so AI serves us, not the other way around,” he said at the announcement.

The centerpiece of LawZero’s technical roadmap is Scientist AI, a system deliberately designed to be non-agentic — meaning it does not pursue goals of its own and is focused purely on identifying objective truths. Rather than competing with frontier models on general task performance, Scientist AI is intended to serve as an oversight layer for other AI systems: auditing their reasoning, flagging misalignment, and accelerating scientific discovery in fields like biology and materials science where rigorous verification matters.

The distinction from commercial labs is deliberate. Where OpenAI, Anthropic, and Google DeepMind build increasingly capable agentic systems, LawZero is betting that the highest-value AI research contribution right now is an independent scientific intelligence that does not have skin in the game.

LawZero, which is incubated at Mila, the Québec AI Institute, will use the funding to establish a Berlin research office, grow its international scientific team, and build sovereign computing infrastructure in Canada through partnerships with Hypertec and 5C data centre providers. The project is expected to create 360 full-time positions in Canada, according to the press release.

Sam Ramadori, former CEO of BrainBox AI, leads the organization day-to-day. The non-profit structure is intentional: Bengio has argued that AI safety research conducted inside for-profit labs faces inherent conflicts of interest, particularly when safety findings could delay or constrain a company’s flagship products.

As reported by BetaKit, the announcement marks a significant escalation of government involvement in the AI safety field, with two G7 nations directly funding an organization whose explicit mandate is to build technical guardrails for the entire AI ecosystem — including systems built by competitors.