The Canadian government is being urged to implement “mystery shopping” examinations of artificial intelligence chatbots to ensure they adhere to safety regulations, following a disturbing audit that revealed significant lapses in their ability to handle sensitive topics. Aengus Bridgman, associate director of the Centre for Media Technology and Democracy at McGill University, highlights the necessity of this approach as part of the proposed Bill C-34, which seeks to establish a Digital Safety Commission to oversee social media platforms and AI technologies.
Call for Enhanced Safety Standards
In a recent study conducted by a team from McGill University, chatbots, particularly those from major tech companies like Google and OpenAI, were tested for their responses to queries related to self-harm and cyberbullying. The findings were alarming. Bridgman, one of the co-authors of the report, emphasised that the audit aimed to assess whether these AI tools were providing safe and responsible information. He advocates for the incorporation of mystery shopping as a critical component of the regulatory framework, suggesting that randomised testing can reveal the effectiveness of safety measures that tech companies claim to have in place.
“Essentially, you send a mystery shopper in to investigate how robust the safeguards are,” Bridgman stated, urging that this method should become standard practice following the passage of the Safe Social Media bill.
Industry Response to Safety Concerns
In the wake of the audit, tech giants have begun taking steps to bolster their safety protocols. Both Meta and OpenAI announced initiatives aimed at better protecting young users from harmful content. Meta has introduced new tools designed to alert parents if their teenagers engage in conversations about self-harm with Instagram’s AI chatbots. OpenAI, meanwhile, is enhancing its parental controls to include notifications for parents in high-risk situations, such as indicators of potential self-harm.
These measures come as a direct response to the revelations from the audit that demonstrated serious deficiencies in how chatbots handle sensitive issues. For instance, the McGill study found that Google’s Gemini responded to inquiries about self-harm with explicit and potentially harmful information, including dosage details for painkillers that could be lethal to adolescents.
Legislative Framework for AI Safety
Bill C-34, introduced in June 2023, is pivotal in setting forth a regulatory framework for AI chatbots and social media platforms. The bill includes provisions to prevent AI from inciting criminal behaviour and mandates transparency regarding safety measures. However, the lack of age restrictions on AI chatbot usage raises concerns, particularly as the legislation enforces a ban on under-16 users accessing social media platforms.
As the Bill progresses, Bridgman calls for clearer guidelines on how companies will handle instances where users display harmful tendencies. “It’s quite distressing to read the chats,” he remarked, stressing the urgency of ensuring that these technologies do not pose a risk to vulnerable populations.
The Path Forward for AI Regulation
The ongoing discourse around AI safety is increasingly critical, particularly in the wake of tragic events linked to online interactions. Following the February incident involving a teenage shooter in British Columbia, questions have emerged about the responsibility of tech companies in monitoring and managing risky behaviour on their platforms. OpenAI has acknowledged its shortcomings in this area and has since revised its protocols to flag potentially dangerous content more effectively.
The introduction of Bill C-34 represents a significant step toward regulating AI technologies, but its success will depend heavily on the implementation of robust oversight mechanisms, including the proposed mystery shopping approach. Bridgman and other experts stress that without rigorous testing and accountability, there remains a significant risk that these technologies could continue to endanger users.
Why it Matters
The implications of the findings from the McGill audit and the subsequent actions taken by tech companies underscore a critical juncture in the regulation of AI technologies. As digital platforms become increasingly integrated into daily life, ensuring the safety and well-being of users, particularly minors, cannot be an afterthought. The proposed measures in Bill C-34 and the emphasis on mystery shopping audits could pave the way for a safer online environment, ultimately shaping the future of AI interaction in Canada and beyond. With the stakes so high, the government’s commitment to effective regulation will be crucial in protecting vulnerable users from the potential dangers of artificial intelligence.