More Canadian companies are touching on artificial intelligence in their financial disclosures, not always positively, according to a new report from the Ontario Securities Commission (OSC).
The regulator published the results of research carried out by its staff to analyze the discussions of AI tools in regulatory filings — specifically, in the management discussion and analysis (MD&A) filings by companies that belong to the S&P/TSX Composite Index over the past 10 years.
That analysis, which was carried out with the help of AI, found that the number of issuers discussing AI in their MD&A increased significantly in their 2024 filings.
In 2015, none of the 225 issuers the OSC reviewed mentioned AI in their disclosure, but in 2024, 32% of companies did. The number of mentions per filing also expanded dramatically, it said.
“The combination of more issuers mentioning AI and these issuers making more mentions per document means that the number of times that issuers mentioned AI increased dramatically over the last decade,” it said.
The OSC also found that the tone of these mentions shifted in companies’ 2024 filings — evolving from largely positive references to more balanced discussions, with companies increasingly pointing out the risks and limitations of AI tools to their businesses.
“In addition to seeing more mentions, the 2024 filing year appears to be a turning point in the overall tone used by issuers to discuss AI,” the report said.
While references to AI were “overwhelmingly positive” before 2024, there was “a significant increase in the number of negative AI mentions,” in 2024, it said.
The regulator also noted that issuers in the finance and information sectors were distinctly more positive about the prospects for AI than companies in other industries.
“This research underscores the evolving landscape of AI in corporate disclosures. As issuers become more attuned to the potential benefits and risks of AI, we see a more nuanced discussion emerging,” said Leslie Byberg, executive vice-president, strategic regulation at the OSC, in a release.
Additionally, the research tested the performance of several large language models (LLMs), concluding that “LLMs can be a valuable tool for securities regulators when conducting research and analysis.”
In particular, regulators are exploring how they can use AI tools in various ways, including to aid their enforcement investigations, for disclosure reviews and in their own research.
“The exploratory analysis presented in this note is one example of an LLM use case, and it has provided promising insights about how LLMs, with appropriate guardrails and human oversight, can enhance the work of regulators,” it said.
The research didn’t look at the adequacy of issuers’ AI disclosure or whether companies were engaging in “AI washing,” or making misleading disclosure about the prospects for AI.