Government 'dependent' on U.S. companies to meet public service AI needs: study

Some organizations within the federal government are “heavily dependent” on American companies for their current or planned AI tools, a new study finds.

With the government turning to artificial intelligence as a key part of its plan to increase efficiency and maintain services while shedding thousands of jobs, a recent study out of the University of Toronto asked how the government’s messaging around the technology stacked up against performance.

“As we are saying those things, are we really putting those things in practice?” asked Dipto Das, a postdoctoral fellow at the University of Toronto and the paper’s lead author.

The research digs into the government’s AI register , a public list of about 400 entries describing current or planned uses of the technology across the federal public service.

For some government organizations, the register only included tools that were developed entirely in-house, while others drew from a mix of internal and external developers.

“One of the things that I was really surprised to see is that there are certain government (organizations)… entirely dependent on third-party developers, like Microsoft, OpenAI, Google,” Das said.

For instance, three out of four of the AI systems listed for the Canadian Radio-television and Telecommunications Commission (CRTC) were developed by Microsoft.

The issue of where an AI tool is developed raises several questions, Das said. How secure is the data? How much control does the government have? And are Canadian taxpayers paying to maintain technological infrastructure on behalf of a third-party company?

“This particular paper is showing our lack of resources and investment in making the Government of Canada technologically sovereign,” he said. “We are seeing a lot of dependency on third parties.”