Canada and Germany have announced plans to support a new approach to artificial intelligence research, focusing on systems designed to be more transparent, reliable and easier to evaluate. During the ALL IN conference in Montréal on September 16, 2026, representatives of the two governments outlined proposed investments in LawZero, an initiative exploring how advanced AI models can be developed with safety considerations incorporated from the beginning. Canada indicated plans to invest CAD 150 million, while Germany announced a proposed contribution of EUR 100 million.
Advertisement
The initiative comes as researchers and technology companies investigate increasingly capable AI systems that can perform complex tasks, generate information and support decisions across different industries. While these developments have created opportunities in medicine, scientific research and business, they have also raised questions about the reliability of automated outputs. Advanced models can produce inaccurate responses, behave unpredictably in unfamiliar situations or make decisions that are difficult for users to interpret. These challenges have encouraged researchers to examine whether safety mechanisms should be integrated into the design of AI systems rather than added only after development.
LawZero’s research direction reflects an effort to reconsider how advanced artificial intelligence is built and evaluated. Instead of measuring progress exclusively through computational performance or task completion, researchers are exploring methods that may improve the ability to understand and verify system behaviour. Such work involves technical challenges, including the evaluation of model accuracy, the interpretation of internal processes and the development of meaningful safety tests. The proposed approach remains a research objective, not evidence that all problems associated with AI reliability have been solved.
