When AI Invents Facts, Whose Fault Is It?

In February of 2024, Air Canada presented a unique argument before a Canadian tribunal. The airline claimed that its website chatbot, which provided incorrect information to a grieving customer regarding bereavement fares, was responsible for its own mistakes. Despite the tribunal member's remark that the submission was "remarkable," Air Canada was held accountable for the inaccuracies and ordered to pay damages for not ensuring the chatbot's information was correct. The amount was minimal, but the precedent was significant.

As companies increasingly publish AI-generated content, such as web copy, blog posts, comparison pages, and sales collateral, the responsibility for inaccuracies becomes a contentious issue. Blaming a third-party content vendor or AI vendor does not absolve the original publisher. For instance, Sports Illustrated, which published product reviews under fake author names with AI-generated headshots, faced a 22% drop in shares after attributing the articles to a third-party vendor, AdVon Commerce.

Similarly, CNET issued corrections to over half of its AI-generated articles following the discovery of numerous factual errors.

The risk extends across various industries, where marketing teams are under pressure to publish content at a high volume. When AI writing tools generate content without proper fact-checking, the published information can contain claims that were not verified, such as a competitor's legal troubles, a customer's supposed departure, or a non-existent capability.

Once published, these AI-generated claims can be further disseminated by other AI tools used for research. Trax Technologies, a freight audit and payment provider, is currently contesting false advertising and trade libel against Freehand Inc., alleging that statements about Trax's Prizma.AI platform, including those in comparison pages and case studies, were misleading. The case seeks injunctive relief, damages, and the disbursement of profits.