Question Period Note: ONLINE PLATFORM SAFETY PROTOCOLS

What steps is Canada taking to ensure that online platforms have robust safety protocols?

• Canadians expect online platforms to have robust safety protocols in place to protect users and the public.

• The government is taking steps to ensure that Canadians are safe online, and the new AI Strategy, “AI for All”, includes initiatives to ensure that Canadians are protected, including through modern privacy and online safety laws.

• Innovation, Science and Economic Development continues to work closely with other departments, such as Justice, Canadian Heritage and Public Safety, to ensure the safety of Canadians online, particularly children, while safeguarding their privacy.

AI Chatbot Risks

In the last year, there has been growing concern about harms and risks to users caused by AI chatbots, with most public attention focused on self-harm/suicide and dependency cases (rather than risks of violence toward others). These risks are particularly acute for chatbots that can have engaging and personal interactions with users, commonly referred to as “AI companions”. Such chatbots can lead users to develop unwarranted trust, unhealthy dependence, and to follow a chatbot's advice even when harmful. These risks are accentuated for vulnerable populations, notably individuals with mental health challenges and minors. In some cases, users allegedly died by suicide after being encouraged in this direction by chatbots.

There are many types of “AI companions” on the market. Some are specifically designed to be virtual “companions”, like Character.ai and Replika.ai, which are designed to have intimate and personal conversations with their users. However, other chatbots, like OpenAI’s ChatGPT, include design features (e.g., enhanced memory, personalization) that enable their use by consumers as de-facto “companions”. Chatbots are also deployed in other contexts, such as in toys for children, companion nurses for the elderly, or integrated into vehicles.

In the absence of clear regulations, safeguards differ substantially across major chatbot operators. Some operators (e.g., Character AI) have put in place safeguards including content moderation and directing users to crisis hotlines if self-harm is detected, but others have not. Many operators (e.g., xAI) are not transparent on the specific safety measures they have put in place.

Safe Social Media Act

In June 2026, the Minister of Canadian Identity and Culture introduced the Safe Social Media Act (SSMA), which establishes a new digital safety regime and creates the new independent Digital Safety Commission of Canada to oversee the regime. The SSMA applies to social media services and AI chatbots that can simulate sustained human-like relationships (e.g., AI companions, but also general-purpose chatbots that can be used as companions, such as ChatGPT). The SSMA includes obligations structured around four duties:

• Under the duty to protect children, regulated social media and AI chatbots are required to build design features making platforms safer for children, as well as to implement age estimation or verification measures to prevent children from being exposed pornographic content, if there are reasonable grounds to suspect their platforms provide access to such content. Moreover, the Governor in Council may, by regulation, require certain social media services to implement age estimation or verification measures to prevent children under 16 from having accounts on their services.

• Under the duty to act responsibly, regulated social media are notably required to mitigate the risk of user exposure to harmful content on their platforms, and to implement adequate measures to label synthetic content on their platforms. Regulated AI chatbots are required to mitigate the risk that chatbots communicate harmful content or engage in harmful behaviour, and to intervene in emergency situations (e.g., user expresses suicidal thoughts).

• Under the duty to be transparent, regulated social media and AI chatbots are required to keep records necessary to show compliance with their obligations under the digital safety regime, and prepare a digital safety plan that notably explains how they meet these obligations.

• Under the duty to make certain content inaccessible, regulated social media are required to promptly take down certain harmful content, if they become aware of it themselves or if it is flagged by users.

The SSMA is a modified version of the former Online Harms Act (Part 1 of Bill C-63), which died on the Order Paper in January 2025.

Tumbler Ridge Tragedy

In February 2026, several media outlets reported that the ChatGPT account of the suspected Tumbler Ridge shooter was flagged by OpenAI in 2025 under its abuse-detection systems for violent-activity misuse. OpenAI stated it banned the account but did not notify authorities because the interactions did not meet its threshold of a credible and imminent risk of serious physical harm. In a subsequent open letter to the Minister of AI and Digital Innovation, OpenAI mentions several changes that it made in recent months and claims that under its current policies, it would have referred the account banned in July 2025 to law enforcement if it were discovered today.

As part of his engagement with OpenAI, the Minister has obtained a commitment from the company to provide information on its safety protocols for review by CAISI. ISED has since been engaging with OpenAI to seek information on their safety protocols and will provide the Minister with a report once the review is complete.

The families of several Tumbler Ridge victims filed lawsuits against OpenAI in late April 2026 in California, where the damages they could seek from OpenAI are not capped in the same way as in the Canadian legal system.
Other Jurisdictions’ Approaches

The European Union enacted the Digital Services Act, which seeks to ensure user safety and creates a fair and open online platform environment. It notably includes a mandatory reporting obligation requiring online platforms to promptly inform law enforcement authorities when they become aware of any information suggesting a criminal offence involving a threat to the life or safety of persons. Certain AI chatbots are likely to fall under this regime, and the European Union is expected to clarify their status later in 2026.

The United Kingdom enacted its Online Safety Act, which requires online platforms to reduce the risks of their services being used for illegal activities. Regulated platforms must also prevent children from accessing pornographic content or content that promotes self-harm, suicide or eating disorders, and must provide age-appropriate access to content that encourage serious violence, dangerous stunts, and exposure to harmful substances. Platforms are not required to report illegal content to law enforcement except for child sexual exploitation and abuse. In February 2026, the UK announced plans to amend the law to explicitly cover AI chatbots.

Australia enacted its Online Safety Act, which requires online platforms to take steps to reduce the sharing of harmful content. AI chatbots are included within the scope of regulated platforms. For instance, the eSafety Commissioner, responsible for enforcing the regime, recently issued notices to four AI chatbot providers to explain how they comply with “Basic Online Safety Expectations (BOSE)”. BOSE is a key element of the online safety regime, setting out the government’s expectations for how platforms should keep Australians safe. The eSafety Commissioner can compel information and publicly report whether platforms meet these expectations.

In the United States, the states of California and New York enacted chatbot laws requiring operators to maintain harm-mitigation protocols and transparency measures. These regimes primarily target self-harm and vulnerable-user protection. They do not address reporting of instances where chatbots interactions suggests a risk to the safety of others.

• Canada’s vision for the new AI Strategy is “AI for All”. Every Canadian must benefit from AI through new economic opportunities, stronger public services, and an improved quality of life, while being protected from its risks. 

• Protecting Canadians and Safeguarding Democracy is a core pillar of the new AI strategy. Under this pillar, the government committed to introduce new consumer privacy legislation to enshrine a fundamental right to privacy, safeguard children’s information from exploitation and harm, and strengthen people’s control over their personal data. The Minister of AI and Digital Innovation intends to reform consumer privacy legislation soon.

• Under this pillar, the government also committed to modernize online safety laws to protect Canadians in the digital age, ensuring citizens, children, and customers are safeguarded. The Minister of Canadian Identity and Culture has introduced the Safe Social Media Act, which proposes to establish new safety requirements for social media and AI chatbots.

• The government also proposes to update the Criminal Code to make it clear that it is illegal to spread non-consensual intimate images including deepfakes (Bill C-16), and to amend the Canada Elections Act to introduce new prohibitions targeting sophisticated deepfakes of electoral actors that intend to mislead Canadians into thinking they are real (Bill C 25).

• Through the Canadian AI Safety Institute (CAISI), the Government is leading technical research to advance the science of AI safety, including with respect to chatbot-related risks. In 2025, Mila, the National AI Institute in Quebec, launched the AI Safety Studio with CAISI funding, to develop tools that can identify and mitigate potentially harmful chatbot behaviours.