AI 治理概况
Canada AI Governance Profile
加拿大人工智能治理的初步观察概况,重点关注联邦公共部门自动化决策问责制、隐私、人工智能战略、政府人工智能登记册、智能体人工智能指导、人工智能安全能力和企业影响。

概况速览
执行摘要
Key signals for Canada AI governance, structured for quick institutional review before the full profile analysis below.
地区
North America
治理模式
Public-sector-led governance
监管成熟度
Developing企业影响
Medium公共部门准备度
Advanced执法成熟度
Developing企业影响
企业应将加拿大视为一个不断发展的治理环境,而非一个仅有一部已颁布的通用人工智能法规的司法管辖区。当前的运营重点包括隐私治理、数据管理、供应商和模型文档、人工监督、安全控制、公共部门采购准备、在相关情况下自愿做出生成式人工智能承诺,以及监测任何新的联邦人工智能立法。向加拿大联邦机构销售产品或与其合作的组织应特别关注自动化决策影响评估的预期和透明度要求。
阅读完整企业影响分析关键机构
- 加拿大财政委员会秘书处
- 加拿大创新、科学和经济发展部
- 加拿大人工智能安全研究所
- 加拿大隐私专员办公室
- 联邦部门实施自动化决策系统
观察清单
我们正在关注什么
可能改变治理格局的后续里程碑、尚未解决的监管问题及实施进展。
- 加拿大是否会在C-27法案/AIDA法案之后推出或推进后续的联邦人工智能立法?
- AIDA政策架构存档或官方替代材料的任何更新
- 加拿大人工智能安全研究所的实施和研究成果
- 关于自动化决策和算法影响评估工具的指令更新
- 联邦隐私法改革和隐私专员办公室关于人工智能的指导意见
- 高级生成式人工智能系统自愿准则的采纳与演进
- 加拿大主权人工智能计算战略和人工智能计算接入基金的实施里程碑
- Government AI Register v2 / expanded coverage by end-2026. Track delivery of the improved, regularly updated replacement described in the September What We Heard report
- 联邦部门的智能体人工智能应用和治理
- 来自金融、卫生、竞争、人权、无障碍、网络安全和劳动力机构的官方行业指导
- Canada–UK AI security partnership: coordinated evaluation work, possible staff exchanges and interoperable standards outputs after the 2026-10-13 renewal
- 2026-11-30 SPIN prioritized-critical-services submission and phased implementation; monitor departmental procurement requirements without extending federal policy to all private firms
- 2027-01-20: OPC Children’s Privacy Code feedback closes; monitor clarification and targeted guidance under existing law
- Government response to OPC’s AI-transparency consultation submission; distinguish the 23 September submission and 23 October announcement from any final adopted rules
- Monitor implementation and any further official updates: Privacy regulators’ resolution on AI in health and health research
来源依据
官方来源依据
执行摘要
加拿大的AI治理架构涵盖联邦公共部门政策、隐私和行业法律、安全和研究机构以及更广泛的AI战略。其最具约束力的运营控制措施集中在联邦行政决策领域:《自动化决策指令》要求对用于制定或支持客户行政决策的自动化系统进行风险评估并采取相应的保障措施,并强制执行算法影响评估。联邦政府还通过公共AI注册系统、《2025-2027年联邦公共服务AI战略》、《智能体AI指南》以及加拿大AI安全研究所等途径,提升透明度和能力。对于私营企业而言,其合规风险仍比欧盟或韩国更为分散;但政府供应商仍应预期政府会对其文档、风险、透明度和保证提出严格的要求。
治理架构
加拿大最清晰、最具约束力的AI治理机制集中在联邦公共部门行政管理领域。《自动化决策指令》及其《算法影响评估》为受监管的联邦自动化决策系统建立了基于风险的控制机制。联邦AI战略、AI注册、智能体AI指南、隐私监管以及加拿大AI安全研究所则进一步增强了透明度、能力和安全性。这些机制不应被推广为适用于私营部门的通用AI法案。
The renewed 13 October UK–Canada partnership supplements domestic institutions with frontier-model evaluation and standards cooperation. It does not create a broad private-sector AI regulatory regime.
主要政策与框架
| Policy | Issuer | 年份 | Status | Summary |
|---|---|---|---|---|
| 关于自动化决策的指令 | Treasury Board of Canada Secretariat | 2019 | Federal public-sector directive - binding within scope | 要求联邦机构通过适当的保障措施、透明度、人为参与、数据治理和监控来评估和管理自动化决策系统。 |
| 算法影响评估 | Government of Canada / Treasury Board of Canada Secretariat | 2026 | Mandatory federal assessment instrument within scope | 支持自动化决策指令的风险评估工具;它不会制定一般的私营部门人工智能法律。 |
| 2025-2027年联邦公共服务人工智能战略 | Government of Canada / Treasury Board of Canada Secretariat | 2025–2027 | Federal public-service strategy | 制定负责任的联邦人工智能应用、劳动力能力、服务改进、治理和公众信任战略。 |
| 加拿大政府人工智能注册 | Treasury Board of Canada Secretariat | 2025–2026 | Public-sector transparency and implementation instrument | The Register implements the federal public-service AI strategy’s transparency commitment. The 2026-09-11 What We Heard consultation report records feedback on the MVP AI Register. Treasury Board says it will use the feedback to develop an improved version replacing the MVP and updated regularly. The report is a non-binding public-service transparency implementation signal, not private-sector AI legislation. |
| 智能体人工智能使用指南 | Government of Canada | 2026 | Official guidance - non-binding outside its administrative context | 基于风险的联邦指导方针,用于考虑智能人工智能的使用、问责制、监督和操作保障措施。 |
| 加拿大人工智能安全研究所 | Innovation, Science and Economic Development Canada | 2026 | AI safety and evaluation institution | 具备先进人工智能安全研究、评估、指导和国际协调的机构能力;而非一般的人工智能监管机构。 |
| C-27号法案/人工智能和数据法案 | Parliament of Canada | 2022 | Archived proposed legislation - not enacted | 这是美国第四十四届议会提出的具有历史意义的联邦人工智能立法提案。C-27号法案最终未能通过,因此不应被视为现行法律。 |
| Direction on Government of Canada Cyber Security Readiness in the Frontier Artificial Intelligence Era — SPIN | Treasury Board of Canada Secretariat | 2026 | Federal administrative security-policy direction — binding within the existing in-scope government policy chain | Effective 1 October, the SPIN reinforces existing government security/digital requirements for organizations covered by section 6 of the Policy on Government Security. It prioritizes critical services, baseline controls and assume-breach recovery. It is not a universal private-sector AI statute. |
| Children’s Privacy Code | Office of the Privacy Commissioner of Canada | 2026 | Regulator guidance interpreting existing privacy law; feedback open, not new legislation | Addresses children’s privacy in digital services, including profiling and AI-driven decisions. Existing law supplies legal duties; guidance and recommendations must not all be treated as statutory commands. |
| Privacy regulators’ resolution on AI in health and health research | Federal, provincial and territorial privacy regulators | 2026 | Non-binding joint regulator resolution | A joint resolution adopted on 15 September and announced on 29 October calls for privacy safeguards, risk assessment and oversight across health AI and research. It is not new legislation. |
政策时间线
里程碑按上次审查时点分组,并保留来源日期的精确程度。
已记录的里程碑
2026-09-11
Government of Canada AI Register — What We Heard
The 2026-09-11 What We Heard consultation report records feedback on the MVP AI Register. Treasury Board says it will use the feedback to develop an improved version replacing the MVP and updated regularly. The report is a non-binding public-service transparency implementation signal, not private-sector AI legislation. Publication date is recorded on the official progress hub; the report page was updated on 2026-09-15.
2026-10-01
Direction on Government of Canada Cyber Security Readiness in the Frontier Artificial Intelligence Era — SPIN
Effective 1 October, the SPIN reinforces existing government security/digital requirements for organizations covered by section 6 of the Policy on Government Security. It prioritizes critical services, baseline controls and assume-breach recovery. It is not a universal private-sector AI statute.
2026-10-13
Canada and the UK renew their AI security partnership
On 13 October 2026 Canada and the UK renewed their 2024 AI safety/security partnership. The Canadian AI Safety Institute and UK AI Security Institute will deepen frontier-model information sharing, evaluation methods, societal resilience and interoperable standards work, exploring staff exchanges and coordinated evaluations. This is bilateral cooperation, not a new horizontal AI law.
2026-10-20
OPC issues Children’s Privacy Code
The Code interprets existing privacy law for children’s data, including age-appropriate notices, consent and profiling/AI-driven decisions. It is regulator guidance, not a newly enacted Children’s AI Act; comments remain open until 20 January 2027.
2026-10-23
OPC announces publication of AI-transparency consultation submission
OPC announced its submission on 23 October; the submission itself is dated 23 September. It addresses generated content, interactions, system information, incidents and agents as consultation input, not newly adopted binding transparency rules.
2026-10-29
Privacy regulators’ resolution on AI in health and health research
A joint resolution adopted on 15 September and announced on 29 October calls for privacy safeguards, risk assessment and oversight across health AI and research. It is not new legislation.
2026-11-09
Paris call for AI safety and security
The joint declaration expresses international cooperation on AI safety and security. It is a non-binding alignment signal rather than new Canadian legislation.
企业影响
企业应将加拿大视为一个不断发展的治理环境,而非一个仅有一部已颁布的通用人工智能法规的司法管辖区。当前的运营重点包括隐私治理、数据管理、供应商和模型文档、人工监督、安全控制、公共部门采购准备、在相关情况下自愿做出生成式人工智能承诺,以及监测任何新的联邦人工智能立法。向加拿大联邦机构销售产品或与其合作的组织应特别关注自动化决策影响评估的预期和透明度要求。
The federal AI Register is evolving from a disclosure commitment toward iterative transparency infrastructure. Its consultation feedback will inform an improved, regularly updated replacement for the MVP. It offers an inventory/disclosure design reference rather than imposing a private-sector AI law.
Suppliers should examine contractual security requirements: departments remain accountable for third-party services and should reflect relevant SPIN requirements in procurement and contract management.
Child-facing AI and EdTech teams should review the Children’s Privacy Code against applicable privacy duties, especially consent, age-appropriate notices and profiling; the Code is interpretive guidance rather than new AI legislation.
Health organisations can compare privacy, procurement and oversight practices with the regulators’ resolution and the laws applicable to them.
观察站解读
加拿大的独特之处在于其制度层次。该政策体系融合了具有约束力的联邦公共部门规则、隐私监管、开放的影响评估工具、自愿性的生成式人工智能承诺、人工智能安全制度建设、智能体人工智能指导以及计算战略。AIDA/C-27法案目前仍处于历史性立法提案阶段,因此实际准备工作取决于现有的行政、隐私、采购、安全和问责控制措施,而非等待一项横向人工智能法律的颁布。
官方资源
| Resource | 来源 | Type | 日期 | Legal force | 为何重要 |
|---|---|---|---|---|---|
包含拟议的AIDA条款的正式法案文本。 | Parliament of Canada | Law | 2022-06-16 | Proposed | 它为加拿大拟议的联邦人工智能法律体系提供了官方立法来源,但其本身并不构成具有约束力的人工智能法律。 |
AIDA框架提案的存档解释性材料。 | Innovation, Science and Economic Development Canada | Guidance | 2023 | Draft | 对于历史解读而言很重要,但由于其已被存档,因此必须注意其法律地位。 |
管理自动化决策系统的联邦公共部门政策文件。 | Treasury Board of Canada Secretariat | Public-sector rule | 2019 | Binding | 这是加拿大联邦政府使用的最清晰、最具约束力的AI治理机制之一。 |
支持自动化决策指令的强制性风险评估工具。 | Government of Canada / Treasury Board of Canada Secretariat | Public-sector rule | 2026-05-28 | Guidance | 将风险分类和缓解预期应用于联邦自动化决策系统。 |
联邦私营部门隐私法。 | Department of Justice Canada | Law | 2000 | Binding | 涉及个人信息的人工智能系统仍然受隐私法约束,不受未来任何人工智能法规的影响。 |
OPC隐私和人工智能资源中心。 | Office of the Privacy Commissioner of Canada | Guidance | Guidance | 展示了加拿大隐私监管机构如何界定与人工智能相关的隐私问题和问责制。 | |
为开发或管理先进生成式人工智能系统的组织提供的自愿承诺。 | Innovation, Science and Economic Development Canada | Framework | 2023 | Voluntary | 在任何具有约束力的横向人工智能法规出台之前,提供一个近期治理信号。 |
加拿大政府人工智能安全研究所专注于高级人工智能风险、研究、指导和国际协调。 | Innovation, Science and Economic Development Canada | Strategy | 2026-06-04 | Not applicable | 在加拿大治理架构中加入人工智能安全制度建设。 |
CAISI 参考的官方网络安全指南,用于安全人工智能系统开发。 | Canadian Centre for Cyber Security | Guidance | Guidance | 将人工智能治理与网络安全控制和安全系统开发联系起来。 | |
加拿大国内人工智能计算资源获取、公共基础设施建设以及对加拿大创新者和研究人员的支持战略。 | Innovation, Science and Economic Development Canada | Strategy | 2026-06-04 | Not applicable | 这表明加拿大的AI治理议程包括计算能力、数据、知识产权和基础设施战略。 |
国家人工智能战略侧重于商业化、标准、人才、研究和应用。 | Innovation, Science and Economic Development Canada | Strategy | 2026-06-04 | Not applicable | 为加拿大的人工智能治理概况提供创新、标准、研究和人才方面的背景信息。 |
官方联邦公共服务人工智能战略,涵盖采用、治理、能力和公众信任。 | Treasury Board of Canada Secretariat | Strategy | 2025 | Not applicable | 展示了联邦人工智能治理与机构能力和劳动力准备情况之间的联系。 |
联邦政府人工智能应用的官方公共部门透明度和实施记录。 | Treasury Board of Canada Secretariat | Public-sector rule | 2025-11-28 | Guidance | 为联邦政府人工智能部署提供可观察的实施和问责信号。 |
联邦政府关于智能体人工智能使用和风险管理的官方指南。 | Government of Canada | Guidance | 2026-05-22 | Guidance | 添加当前智能体人工智能实施信号,但不会过度强调指导意见作为具有约束力的法律。 |
The 2026-09-11 What We Heard consultation report records feedback on the MVP AI Register. Treasury Board says it will use the feedback to develop an improved version replacing the MVP and updated regularly. The report is a non-binding public-service transparency implementation signal, not private-sector AI legislation. Publication date is recorded on the official progress hub; the report page was updated on 2026-09-15. | Treasury Board of Canada Secretariat | 报告 | 2026-09-11 | Not applicable | Practical reference for AI inventories and disclosure design; not a private-sector statute. |
Official implementation hub confirming publication of the AI Register consultation report on 2026-09-11. | Treasury Board of Canada Secretariat | 报告 | 2026-09-15 | Not applicable | Distinguishes the publication date from the report page’s later update date. |
On 13 October 2026 Canada and the UK renewed their 2024 AI safety/security partnership. The Canadian AI Safety Institute and UK AI Security Institute will deepen frontier-model information sharing, evaluation methods, societal resilience and interoperable standards work, exploring staff exchanges and coordinated evaluations. This is bilateral cooperation, not a new horizontal AI law. | ISED Canada / UK Government | 报告 | 2026-10-13 | Not applicable | One bilateral event referenced by two profiles; cooperation does not replace domestic legal obligations. |
Official counterpart statement on the same bilateral partnership, not a second instrument. | UK Cabinet Office | 报告 | 2026-10-13 | Not applicable | Corroborates the shared event and institutional anchors. |
Effective 1 October, the SPIN reinforces existing government security/digital requirements for organizations covered by section 6 of the Policy on Government Security. It prioritizes critical services, baseline controls and assume-breach recovery. It is not a universal private-sector AI statute. | Treasury Board of Canada Secretariat | Public-sector rule | 2026-10-01 | Binding | Suppliers should examine contractual security requirements: departments remain accountable for third-party services and should reflect relevant SPIN requirements in procurement and contract management. |
The Code interprets existing privacy law for children’s data, including age-appropriate notices, consent and profiling/AI-driven decisions. It is regulator guidance, not a newly enacted Children’s AI Act; comments remain open until 20 January 2027. | Office of the Privacy Commissioner of Canada | Guidance | 2026-10-20 | Guidance | Maintains the distinction between this development and existing binding AI duties. |
OPC announced its submission on 23 October; the submission itself is dated 23 September. It addresses generated content, interactions, system information, incidents and agents as consultation input, not newly adopted binding transparency rules. | Office of the Privacy Commissioner of Canada | 报告 | 2026-10-23 | Not applicable | Maintains the distinction between this development and existing binding AI duties. |
Submission dated 23 September, publicly announced 23 October 2026. Regulator consultation position, not a final government rule. | Office of the Privacy Commissioner of Canada | 报告 | 2026-09-23 | Not applicable | Preserves submission date separately from public announcement and any future policy adoption. |
A joint resolution adopted on 15 September and announced on 29 October calls for privacy safeguards, risk assessment and oversight across health AI and research. It is not new legislation. | Federal, provincial and territorial privacy regulators | Guidance | 2026-10-29 | Guidance | Non-binding joint regulator resolution |
The joint declaration expresses international cooperation on AI safety and security. It is a non-binding alignment signal rather than new Canadian legislation. | Government of Canada and international signatories | Strategy | 2026-11-09 | Voluntary | Official evidence for the stated development and its legal status. |