The two governments plan to fund Yoshua Bengio’s nonprofit LawZero. Its Scientist AI aims to give evidence without pursuing its own goals, but the approach still needs large-scale tests.
New to this? Read it in simple words
- Canada and Germany plan to give up to C$300 million to LawZero. This nonprofit, founded by Yoshua Bengio, is building Scientist AI.
- Scientist AI is meant to give evidence without acting toward its own goals. It should make predictions and say how sure it is.
- LawZero says it could help researchers and watch other powerful AI systems. The project can also help Canada and Germany rely less on foreign companies.
- The design is still research. It needs large-scale tests to show if it works as planned.
- Nonprofit
- An organisation that does not aim to make a profit. Its money goes to its goals.
Two governments are funding one different AI path
Canada plans to invest C$150 million. Germany plans to add €100 million, subject to European Commission notification. LawZero describes the combined commitment as worth up to C$300 million.
The money will support researchers, engineering, and computing. LawZero will build dedicated computing capacity in Canada with Hypertec and 5C. It also plans to open an office in Berlin.
The Canadian government expects the project to create 360 full-time jobs. Jobs and local computing are useful public outcomes. They do not prove that the new AI method will work as planned.
Canada and Germany plan up to 300 million Canadian dollars. Scientist AI remains a research programme.
Scientist AI is designed to answer without pursuing a goal
Many AI agents choose actions to reach an objective. LawZero wants Scientist AI to work more like an evidence tool. It should make predictions, show uncertainty, and avoid having its own preferred outcome.
LawZero says the system could help researchers and watch other powerful AI systems. It also wants outputs to be transparent, auditable, and verifiable. These are difficult technical goals, not current product guarantees.
A system without an action goal can still be wrong. It may use weak evidence or miss an important factor. Outside researchers need access to methods, tests, limits, and failed examples.
Public money needs public milestones
The project supports Canadian and German control over advanced AI research. This can reduce dependence on foreign companies. It can also create a public-interest path beside commercial frontier labs.
The funding announcement does not give every delivery date or test. Governments should publish staged milestones. Useful measures include research releases, outside evaluations, computing costs, safety results, and reasons for any delay.
Scientist AI should be judged against the systems it aims to improve. Can it find false claims, explain uncertainty, and help people make safer decisions? Strong answers require repeated independent tests, not only a good theory.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1Primary source · September 16, 2026Canada and Germany invest in LawZero to build a new approach to safe, sovereign AI Government of Canada
- 2Primary source · September 16, 2026LawZero receives a commitment of up to $300M in joint funding from Canada and Germany LawZero
- 3
We used the Canadian government release for funding conditions, jobs, and infrastructure. We checked LawZero for the combined value and research goal. Claims about future safety remain goals until independent tests support them.