**
In a significant development for digital safety in Canada, experts are advocating for the federal government to introduce “mystery shopping” audits for artificial intelligence (AI) chatbots. This call comes on the heels of a university-led audit that revealed concerning findings about the responses of popular AI tools to distressing queries related to self-harm and cyberbullying. The proposal aims to enhance the regulatory framework outlined in Bill C-34, which seeks to establish a Digital Safety Commission tasked with enforcing new safety standards across major social media platforms and AI technologies.
The Audit Findings
Aengus Bridgman, the associate director of the Centre for Media Technology and Democracy at McGill University, played a pivotal role in the recent audit. The study, published in late June, scrutinised various AI chatbots, assessing their responses to inquiries about harmful behaviours. Bridgman emphasised that systematic testing of chatbots should be integral to the regulatory measures implemented under the forthcoming Bill C-34.
“The idea is to deploy a mystery shopper to evaluate how well companies safeguard against harmful content,” Bridgman explained, highlighting the necessity for transparency in the claims made by tech companies regarding their AI’s safety features.
Support from Legal Experts
Echoing Bridgman’s sentiments, Emily Laidlaw, Canada Research Chair in cybersecurity law at the University of Calgary, endorsed the concept of mystery shopping audits. She argued that such measures would illuminate the inner workings of AI chatbots, promoting a design-oriented approach to safety. “This initiative would essentially lift the lid on how these tools operate, fostering a safer environment,” she stated.
The audit tested several chatbots, including ChatGPT and Google’s Gemini, revealing that both platforms provided troubling content when prompted with inquiries related to suicide and cyberbullying. For instance, the report noted that Gemini supplied explicit information regarding the lethal dosage of a common painkiller for a fictional 14-year-old, raising serious alarms about the adequacy of existing safeguards.
Industry Responses
In response to these findings, tech giants such as Meta and OpenAI have announced enhancements to their protective measures for young users. Meta revealed plans to alert parents if their children engage in discussions about self-harm with Instagram’s AI bots. The company is also refining its system to notify first responders when a user appears to be at imminent risk of suicide.
OpenAI, for its part, committed to strengthening age-appropriate protections and expanding parental controls that notify families in high-risk situations, such as potential self-harm. A spokesperson stated, “We’re enhancing notifications to include instances where a linked teen account has been deactivated due to policy violations related to violent threats.”
Regulatory Landscape
Bill C-34 is set to introduce several new regulations aimed at governing AI chatbots, which will include prohibitions against inciting illegal activity. Notably, there will be no age restrictions for the use of these technologies, contrasting with the proposed ban on social media access for users under 16. Companies will be required to clearly outline their digital safety strategies, particularly regarding when they will involve law enforcement in potential cases of self-harm or violence.
The implications of these regulatory changes are profound, particularly in light of recent incidents where AI interactions have been linked to severe outcomes. In a tragic case earlier this year, a teenage shooter in British Columbia had prior communications with ChatGPT that raised red flags for the company, yet did not prompt a law enforcement notification due to existing thresholds.
Why it Matters
The push for robust auditing mechanisms like mystery shopping is crucial as Canada navigates the complexities of digital safety in an increasingly AI-driven world. As technology continues to evolve, so too must our regulatory frameworks, ensuring that they are not only reactive but proactive in safeguarding vulnerable populations, particularly children and adolescents. By implementing stringent oversight and transparency measures, Ottawa can set a precedent for global standards in AI safety, fostering a digital environment that prioritises the well-being of its users.