Artificial Intelligence (AI) maintains a reputation of fear and fascination as it expands in accessibility, capability, and reliability. ChatGPT, and all of its increasing use cases, continues to command the public’s attention. Dall-E and other AI image generators (see our team’s take on navigating AI generated images here) have the world questioning the Pope’s latest attire while AI-music generators are making Drake’s latest hits, without Drake.
While the realities of AI may be somewhat new to the general public, its presence may not be as sudden as we imagine. In June of 2022, in anticipation of advancements breaking through in AI, the Minister of Innovation, Science and Industry introduced Bill C-27 which notably includes the Artificial Intelligence and Data Act (AIDA).
AIDA was proposed with the intentions of regulating AI development and use across Canada. It attempts to be a proactive, principle-based approach to the governance of AI capabilities. It will not take a rights-based approach to AI regulations, meaning, unlike privacy legislation in the past, this law will not create any new rights for an individual. This legislation will protect an individual from harm to themselves (physical or psychological) or their property, and economic loss, thus upholding existing individual rights in relation to AI.
The Bill will establish a standard of transparency and responsibility for those developing and commercializing AI tools. This will likely include data protection standards and impact assessments that will then require those responsible for the AI to develop and monitor systems of risk mitigation. Those that meet certain risk criteria, and are asked to develop methods of risk mitigation, will be labelled “High Impact”. The term High Impact has not gone through the necessary consultation phase to accurately define it, however, the expectation is that it will be determined based on:
- evidence of risk of harm to health and safety;
- severity of potential harms;
- scale of use;
- system opt out ability; and
- risk regulation under other laws and legislation
Faskens has affirmed that AIDA will focus on AI in screening systems for services or employment, biometric systems used for identification and applications like AI-powered online content recommendation, autonomous vehicles, and health related tech. Notably, this does not include open-source software.
OpenAI has also commented on the necessary regulation of, what they call, “superintelligence”. Notably, the platform has shared sentiments in line with what AIDA suggests, claiming that AI with capabilities above a certain threshold need to be subject to system inspections, audits, compliance and safety tests, and more. However, they do also emphasize the necessity of allowing companies to develop AI models below a certain capability threshold.
AIDA is still making its way through government channels and is not yet in effect. However, if Bill C-27 is approved and implemented, what will that mean for businesses building themselves on and around AI in Canada?
A recent companion document to AIDA has been a topic of discussion in Parliament as a means of addressing that very question. Innovation, Science and Economic Development Canada published the document which attempts to reassure the concerned public seeking regulation and the AI stakeholders anxiously awaiting regulatory details. In summary, the companion document states that AIDA intends to address the impact of AI whilst maintaining the flexibility required for positive growth in the tech industry, providing an outline of regulations and criteria that will be flagged.
While Canada was an early starter, standing alone with the European Union as the only two governments moving through official channels for AI legislation at the time that AIDA was proposed, it will not be until at least 2025 that AIDA will come into force. AIDA also appears to be simpler than the EU legislation, meaning that the Bill C-27 does not appear to be heavy handed with regulations. On the other hand, the simplicity of AIDA threatens to introduce uncertainty.
Definitions, regulation details, and the fine print of this Bill has yet to be written – leaving many unsure of what AI’s future in Canada looks like. Whatever may be in store, it seems to be true that ChatGPT, and others like it, are here to stay. As OpenAI has put it, “we can have a dramatically more prosperous future; but we have to manage risk to get there”.