Question Period Note: VOLUNTARY CODE OF CONDUCT FOR ADVANCED GENERATIVE ARTIFICIAL INTELLIGENCE SYSTEMS

How is the voluntary code for artificial intelligence helping reassure Canadians that AI will be adopted in a safe manner?

• The Government of Canada is committed to deploying and developing artificial intelligence in a safe and secure way for all Canadians.

• Canada’s Voluntary Code of Conduct for Advanced Generative Artificial Intelligence Systems sets out practices that firms can implement immediately to address risks associated with advanced generative AI, in advance of the Artificial Intelligence and Data Act becoming law.

• The Code was developed in close collaboration with industry and other stakeholders, and already has more than 30 committed organizations, with more steadily showing interest.

The launch of OpenAI’s ChatGPT in November 2022 led to growing concern internationally on the risks posed by advanced generative AI systems. Recognizing the urgent need for guardrails on generative Artificial Intelligence (AI), the Government of Canada launched a consultation on potential elements of a voluntary code of conduct for generative AI systems in Summer 2023. The government had some great discussions with innovators, researchers, and civil society stakeholders about what firms need to do to build trust in their activities, and how the government can make sure that the code is practical and implementable.

As a result, the government announced the Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems (the Code) in September 2023. The Code identifies measures that organizations can apply to their operations when they are developing and managing generative AI systems with general-purpose capabilities and provides a critical bridge between now and the coming into force of Canada’s proposed regulatory regime under the Artificial Intelligence and Data Act (AIDA). To date, more than 30 organizations working with generative AI have signed onto the Code and committed to respect its principles.

The Code takes a risk-based approach by segmenting responsibilities based on where firms sit in the value chain and how the system is deployed. Firms developing and managing the operations of advanced generative systems both have important and complementary roles. Developers are responsible for ensuring that due diligence has been done in terms of potential adverse impacts prior to the release of new versions of their systems. After deployment, firms managing the operations of such systems are responsible for monitoring systems in operation for inappropriate or malicious use with the potential to cause adverse impacts.  Managers and developers are expected to cooperate in order to ensure that adverse impacts are addressed by the firm who has the ability to do so.  

The Code is also aligned with the six principles laid out in the AIDA companion document. These principles are in-line with international norms on AI-systems governance and focus on outcomes needed to build trust with Canadians, including: 

• Accountability – Firms understand their role with regard to the systems they develop or manage, put in place appropriate risk management systems, and cooperate with other firms as needed to avoid gaps.  

• Safety – Systems are subject to risk assessments, and mitigations needed to ensure safe operation are put in place prior to deployment and re-evaluated as needed.  

• Fairness and Equity – Potential impacts with regard to fairness and equity are assessed and addressed at different phases of development and deployment of the systems.  

• Human Oversight and Monitoring – Systems include appropriate mechanisms for human oversight, and use is monitored after deployment. Information is shared and updates are implemented as needed to address any risks that materialize.   

• Validity and Robustness – Systems operate as intended and their behaviour in response to the range of tasks or situations to which they are likely to be exposed is understood.  

• Transparency – Sufficient information is published to allow consumers to make informed decisions and for experts to evaluate whether risks have been adequately addressed.  

Signatories to the Code commit to undertake specific measures under each of the six principles, depending on whether they are a developer or manager of an AI system. They also commit to supporting the ongoing development of a robust, responsible AI ecosystem in Canada, for instance by contributing to the development and application of standards, sharing information and best practices with other members of the AI ecosystem, and collaborating with researchers to support public awareness and accountability. Finally, signatories commit to developing and deploying AI systems consistent with accessibility, human rights, and environmental sustainability, and working collaboratively to harness the potential of AI to address pressing global issues such as climate change.

• Signatories to the Voluntary Code of Conduct for Advanced Generative Artificial Intelligence (AI) Systems commit to adopting measures related to the development and management of operations of advanced generative AI systems to address and mitigate risks associated with such systems.

• The measures are intended to be targeted, principles-based, and to be applied by firms in a manner that is proportionate to risk.  

• While the measures prescribed in the Code are specific to the generative AI context, the principles underpinning them are broadly applicable to a range of AI systems and can be readily adopted by actors across Canada’s AI ecosystem. 

• To date, more than 30 organizations working with generative AI have signed onto the Code and committed to respect its principles, including Mastercard, TELUS, and Cohere. This number continues to grow with more signatories anticipated this fall.

• This steady growth of signatories to the Code is a testament to Canadian firms commitment to implementing responsible AI.