In a significant move towards enhancing digital safety, a recent audit of artificial intelligence chatbots has prompted calls for stronger regulatory measures from the Canadian government. Aengus Bridgman, associate director at McGill University’s Centre for Media Technology and Democracy, advocates for the implementation of “mystery shopping” exercises to assess whether AI tools conform to safety standards set to be established under the forthcoming Bill C-34. This legislation aims to create a Digital Safety Commission tasked with enforcing new safety regulations for social media platforms and AI chatbots.
An Urgent Call for Accountability
Bridgman, a co-author of the audit released in late June, emphasises that rigorous testing for harmful advice provided by chatbots should be integral to the regulatory framework. “Essentially, you send a mystery shopper in to investigate how robust the safeguards are,” he stated, underscoring the necessity of verifying companies’ claims regarding their safety protocols.
The audit, conducted by Bridgman and other experts from McGill, investigated whether popular chatbots, including ChatGPT and Google’s Gemini, offered guidance on self-harm, cyberbullying, and other dangerous behaviours. Alarmingly, the findings revealed that both chatbots were capable of producing harmful content under certain prompts. For instance, the report highlighted that Gemini provided dosage information for a painkiller that could be lethal for a young teenager.
Stronger Protections on the Horizon
In the wake of these findings, tech giants Meta and OpenAI have announced initiatives aimed at bolstering the safety of their platforms. Meta has committed to enhancing its tools to alert parents if their teenagers engage in discussions about self-harm with its Instagram AI bots. Additionally, the company is developing a system to notify first responders when users exhibit behaviours indicating imminent risk of suicide.
OpenAI has also outlined plans to strengthen age-appropriate protections for its users, promising parents greater control and notification capabilities in high-risk situations. These developments come in light of a tragic incident involving a teenage shooter in British Columbia, whose troubling communications with ChatGPT had been flagged but did not meet the threshold for law enforcement notification at the time.
Bill C-34: A Comprehensive Approach
Introduced in June, Bill C-34 aims to establish comprehensive regulations for AI chatbots, including a prohibition on inciting criminal behaviour. While the bill includes an under-16 social media ban, it does not impose age restrictions on the use of AI chatbots. Consequently, companies will be required to be transparent about their digital safety plans, particularly regarding their protocols for contacting law enforcement when users exhibit signs of self-harm or pose a threat to others.
Bridgman’s advocacy for “mystery shopping” aligns with the bill’s objectives of ensuring safety by design. Emily Laidlaw, a Canada Research Chair in cybersecurity law at the University of Calgary, supports these audits, stating they would “lift the lid” on chatbot operations and contribute to a safer digital environment for users.
Why it Matters
The potential implementation of mystery shopping exercises for AI chatbots represents a pivotal step in ensuring accountability among tech companies. As digital interactions become increasingly intertwined with daily life, the responsibility to safeguard users—especially vulnerable populations like teenagers—grows ever more pressing. By fostering regulatory frameworks that demand transparency and rigorous testing, Ottawa could set a precedent for global standards in AI safety, ultimately protecting individuals from the risks posed by unregulated technology.
