The Last Word: AI in 2020

By Adam Allouba, Partner, Dentons

Every new technology brings new risks and opportunities, and artificial intelligence is no exception. AI forces us to consider both optimistic and pessimistic scenarios, previously reserved for science fiction. While it will take an army of thinkers to unravel the cultural, sociological, economic and other implications of AI, as a lawyer I can share a few thoughts on the legal aspects.

The pros and cons of AI are often two sides of the same coin. For example, AI can analyze massive quantities of data and derive a pattern where the human brain cannot. As a result, it improves our ability to make data-driven decisions, and even introduce evidence-based decision-making to fields that historically relied only on human intuition. In areas ranging from human resources to systems controls to medical diagnosis, AI can help facilitate better decisions, and more efficiently.

That opportunity, however, is not without its challenges. Given that AI runs on data – much of it personal information – it is worth focusing on Canadian privacy laws. Originally written decades ago, they were designed for an era when the internet was embryonic, and “privacy concerns” referred to junk mail and telemarketers. The cornerstone of those laws is informed consent, which could realistically be obtained when the major concern was, for example, allowing a magazine publisher to share your contact information with a direct marketer. However, in a world with endless privacy policies that most of us ignore when creating online accounts, does clicking “Accept” really mean that you have legitimately agreed to anything?

There have been significant modernizing reforms of privacy laws in other jurisdictions, such as Europe (you may have heard of the General Data Protection Regulation or GDPR). And while we may not want to imitate them in every respect, it is worth noting that even on a continent where privacy concerns run high, the law allows an individual’s personal data to be processed without consent if necessary for an organization’s “legitimate interests.”

Given the impossibility of obtaining informed consent for every collection and use of personal information in a world of ubiquitous AI, it might be best for the consent requirement to be limited to data that truly matters, such as financial, health-related, biometric, geolocation and other particularly sensitive information. Such an approach would mean no longer devoting scarce regulatory resources to policing the treatment of more mundane data such as height, weight and shoe size.

Depending on the specific use, AI can also raise profound ethical issues that can lead to legal problems. For example, using AI to scan a playlist and recommend additional songs give little cause for concern. Using it to evaluate credit applications, however, is another story. An unintentionally biased AI could end up rejecting loan applications based on prohibited demographic characteristics such as gender or ethnicity, creating significant legal risk. Before deploying an AI system, organizations should study one of the many available ethical frameworks (I personally recommend the Montreal Declaration), and integrate its recommendations, so as to mitigate the danger of civil lawsuits and regulatory action.

AI’s enormous potential to revolutionize society makes it essential to think carefully about how to unlock its rewards without giving rise to its risks. The law must evolve in a way that reflects the technology, and the social change it will bring. In the meantime, organizations should     seek the expert advice necessary to navigate the uncertain legal and regulatory environment of artificial intelligence.

Adam Allouba is a Montreal-based partner with Dentons, the world’s largest law firm. He is the Montreal lead for the firm’s Transformative Technologies and Data Strategy group, working with startups, SMEs and large businesses. Adam is a member of the AI Advisory Board of Dawson College, has taught a college-level class on law and ethics in AI, and speaks and writes regularly on topics of interest to the AI community.

 Audio Version

Female Voice Audio