Officials from four of the biggest tech companies in the world — Amazon, Google, Microsoft and Meta — largely offered polite criticism of the country’s proposed artificial intelligence law to Canadian parliamentarians for over an hour at a hearing Wednesday.
Several agreed the Artificial Intelligence and Data Act (AIDA) legislation should be passed quickly, but with certain provisions more clearly defined, and allowing more rules to be set in regulations than in the law so it will be flexible.
And then it was like a mask dropped.
It came after Microsoft Canada’s Amanda Craig, the company’s senior director of public policy in the office of responsible AI, gave a lengthy explanation of things that could be improved.
Then Rachel Curran, head of public policy for Meta Platforms in Canada, was asked to comment.
“I think Amanda is being very diplomatic,” she told the House of Commons industry committee.
“AIDA in a number of respects goes well beyond the most stringent proposal out there internationally, which is the EU Act — which is already the subject of a lot of debate amongst [European Union] member states. It doesn’t have the support of countries like France, for instance, who want to make sure their own domestic industry is given a chance to flourish. So AIDA has created a standard that doesn’t exist anywhere else in the world.”
If AIDA passes as is, Meta could meet its requirements for regulating AI systems, she said. But, Curran added “the compliance costs are incredibly high. Would that mean certain [Meta] products would not be launched in Canada? Maybe. But all of us work for companies that are able to meet very high [regulatory] thresholds because we have the resources and money to do that.” On the other hand, “it’s going to have a significant impact on the Canadian AI industry and on innovation in Canada.
“Canada should make sure it aligns itself [in AI legislation] with other jurisdictions. We’re a relatively small market. The EU is setting a benchmark that is world-leading. We should at the very least not exceed that.”
Passing AI legislation fast is important, Curran said, if only to maintain public trust in artificial intelligence applications. But, she added, it’s also important to get the legislation right. AIDA is “a pretty good bill,” she said at one point. “It’s just a question of whether we can get the good details right.”
But, at another point, she said this: “Our global AI research is based in Montreal. The wrong regulatory framework, over-reach, or over-regulation by the government would drive that kind of activity out of the country. And I would hate to see that because we are leaders in AI research.”
Nicole Foster, director of Amazon Web Services’ global artificial intelligence and Canada public policy, agreed with Curran on the need to not get ahead of other countries. At the very least, she added, MPs should hear the opinions on the impact of AIDA from Canadian firms who are, or will be, using AI applications.
AIDA is part of Bill-C-26, which includes a proposed Consumer Privacy Protection Act (CPPA).
AIDA would oversee three classes of AI systems — high-impact, general impact, and machine learning systems — used in areas such as employment, providing services to an individual, processing biometric information for identification, moderating content on a search engine or social media platform, and more. It would be illegal to deploy an AI system likely to cause serious physical, psychological, or economic harm to an individual. Persons responsible for high-impact systems would have to establish measures to identify, assess, and mitigate the risks of harm or biased output that could result from the use of the AI system.
In its initial version, AIDA left a lot of grey areas — such as defining a “high impact” AI system — to be filled in by regulations proclaimed by the government. Those regulations would be set after consultations with experts, including the tech industry. The approach would allow flexibility to meet the challenges of fast-moving AI technology. But some worried about passing vague legislation. So Innovation Minister Champagne sent the committee a letter outlining amendments and clarifications the government is willing to make to the legislation.
However, Curran believes the original version — leaving a lot to regulations — was better. “The minister’s proposed amendments, if accepted by this committee, are going to box Canada into a regulatory framework that may look very different than the one that emerges from international discussions,” she said.
In addition, Curran objected to AIDA covering the moderation of content on social media platforms like Meta. Controls over social media would be better put in the government’s long-promised online harms bill, she said.
Several of the witnesses Wednesday also objected to AIDA’s proposal to give a new AI and Data Commissioner power to enter a firm’s premises, access systems, copy data, and conduct testing of AI systems if the commissioner has reasonable grounds to believe that an organization has contravened or is likely to contravene their obligations under AIDA.
There were also objections over the proposal that employees could be convicted of a criminal offence for mishandling personal data from an AI system. And there were suggestions that some enforcement of AIDA should be put in the hands of regulators that already look after specific industries.
“I’m a bit confused,” Bloc Quebecois MP Jean-Denis Garon admitted late in the session. “You say Parliament needs to regulate AI,” he said to the four witnesses, “then you say C-27 is trying to cover too much, and one suggests it may be better to regulate AI by amending existing legislation covering regulated sectors.” That, he said would require “un-ending legislative work …and bottom line, we’d end up with no regulation … Is this your way of telling us you don’t want regulation?”
“We support good regulatory frameworks,” replied AWS’s Foster. “All of us do.” But if acting fast is the issue, potentially the government can move faster by amending existing legislation.
The U.K. has decided for the time being to let regulators such as the Information Commissioner and the Competition and Market Authority to issue regulations for AI oversight.