AI 治理概况
Australia AI Governance Profile
澳大利亚人工智能治理的初步观察概况,涵盖联邦公共部门控制、国家人工智能计划、人工智能安全能力、自动化决策的隐私透明度规则以及拟议的高风险防护措施的现状。

概况速览
执行摘要
Key signals for Australia AI governance, structured for quick institutional review before the full profile analysis below.
地区
Asia-Pacific
治理模式
Innovation-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
企业不应将封闭式强制性保障措施咨询视为现行法律。近期管控措施应侧重于现有的隐私、消费者、行业和安全义务;在2026年12月10日之前,确保重大自动化决策的透明度和记录保存;模型和数据文档;人工审核和升级机制;测试和保障;采购和供应商管控;以及监测未来联邦或行业特定规则的出台。
阅读完整企业影响分析关键机构
- 工业、科学和资源部
- 数字转型机构
- 澳大利亚信息专员办公室
- 澳大利亚人工智能安全研究所
- 澳大利亚竞争与消费者委员会
- 行业监管机构和联邦部门
- AI Review Committee — non-binding public-sector advisory body
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。
- Preparation for the enacted APP automated-decision transparency requirements commencing 2026-12-10, using OAIC’s 2026-09-30 guidance and scope conditions
- 政府是否会重新审视强制性高风险人工智能防护措施或特定行业的强制性规则
- 政府负责任地使用人工智能政策的实施更新
- 澳大利亚人工智能安全研究所的测试、评估和标准输出
- OAIC关于人工智能、透明度、隐私和自动化决策的指导和执法
- ACCC和行业监管机构就人工智能赋能的消费者、竞争、健康、金融或就业实践采取的行动
- 国家人工智能计划实施、劳动力能力和公共部门采购信号
- Outcome of OAIC’s 2026-10-07 investigation into the maker of the HeyCyan smart-glasses phone app; monitor findings or further action without assuming a violation or penalty has been established
- AI Review Committee advice, agency uptake and annual review of operating arrangements; committee advice remains non-binding
- Monitor implementation and any further official updates: Australian Government AI Reference Architecture
- Monitor implementation and any further official updates: Fair Work Commission requirements for GenAI in cases
- Proposed frontier AI standards legislation, regulator design, consultation and any eventual commencement
- Implementation of Rapid Review recommendations and APRA AI-risk supervisory engagement
执行摘要
澳大利亚目前没有统一的人工智能法案或强制性的高风险人工智能监管机制。现有的架构结合了国家人工智能计划、政府负责任使用人工智能的强制性政策、现有的隐私和消费者保护措施、澳大利亚人工智能安全研究所,以及针对特定行业的监管工作。
对于公共机构而言,联邦政府的政策在其适用范围内建立了负责任的人工智能使用强制性内部控制机制。对于私营企业而言,近期主要的法律进展是针对某些自动化决策的隐私透明度要求已正式生效,该要求计划于2026年12月10日开始实施。拟议的强制性保障措施咨询已经结束,但政府目前尚未推进该项广泛的制度;因此,它应继续作为一项已结束或暂缓实施的政策路径,而非现行法律。
因此,澳大利亚的治理信号是执行导向型和适应型的。企业准备工作应侧重于隐私、透明度、记录保存、消费者保护、行业义务、安全测试、采购控制以及对未来立法或监管行动的监测。
On 10 November 2026 the government announced proposed mandatory protections for frontier AI and a targeted regulator. These remain proposals; the previously discontinued broad high-risk guardrails proposal remains part of the policy history.
治理架构
澳大利亚的人工智能治理架构由联邦行政政策、隐私和消费者法律、行业监管、安全和评估基础设施以及国家战略层层叠加而成。《政府负责任使用人工智能政策》是联邦公共部门的一项强制性管控措施。《国家人工智能计划》侧重于战略层面而非强制性规定;澳大利亚人工智能安全研究所提供能力和测试基础设施,而非一般的监管权限;即将修订的《隐私法》透明度修正案针对某些自动化决策提出了有针对性的法律要求。此前强制性的高风险防护措施机制尚未正式实施。
The AI Review Committee adds expert, non-binding pre-deployment advice on selected agency AI uses and review of serious-incident responses. DTA reported the 26 June inaugural meeting on 7 October; agencies retain accountability.
主要政策与框架
| Policy | Issuer | 年份 | Status | Summary |
|---|---|---|---|---|
| 国家人工智能计划 | Australian Government / Department of Industry, Science and Resources | 2025 | National strategy | 国家人工智能能力、应用、安全、投资、技能和公共利益战略。这是一项战略性政策,而非具有约束力的独立人工智能义务来源。 |
| 高风险人工智能的强制性防护措施——磋商和政策现状 | Australian Government / Department of Industry, Science and Resources | 2024–2026 | Closed policy pathway — government not proceeding with broad mandatory guardrails at this time | 官方咨询和状态页面记录了拟议的强制性高风险人工智能防护措施的政策制定过程。该提案目前不具有法律约束力,不应被视为已颁布的制度。 |
| 政府负责任地使用人工智能政策 v2.0 | Australian Government / Digital Transformation Agency | 2025 | Mandatory Commonwealth public-sector policy within scope | 联邦政府使用人工智能的强制性政策控制,包括负责任的使用、风险管理、问责制、透明度和保证期望。 |
| 澳大利亚人工智能安全研究所 | Australian Government / Department of Industry, Science and Resources | 2025–2026 | AI safety and evaluation institution | 政府具备人工智能安全研究、评估、测试、标准制定和国际合作的能力;但它不是一般的人工智能监管机构。 |
| 《隐私法》自动化决策透明度修正案 | Office of the Australian Information Commissioner / Australian Government | 2024–2026 | Enacted privacy-law amendments — scheduled to commence 2026-12-10 | The Privacy and Other Legislation Amendment Act 2024 enacted APP privacy-policy transparency requirements commencing 2026-12-10. They apply where an APP entity arranges for a computer program to make, or do something substantially and directly related to making, a decision that could reasonably significantly affect an individual’s rights or interests, and uses personal information about that individual in the program. As of 2026-10-03 these requirements are enacted but not yet applicable. |
| Australian Government AI Reference Architecture | Digital Transformation Agency | 2026 | Non-mandatory public-sector reference architecture | DTA announced a reference architecture for public agencies designing and governing AI systems, including assurance and human oversight. The architecture is guidance, not a national AI statute. |
| Fair Work Commission requirements for GenAI in cases | Fair Work Commission | 2026 | Case-procedure requirements | Participants using GenAI to prepare Commission case documents must disclose its use and check accuracy and relevance; witness statements must reflect personal knowledge and truthful words. The requirements concern Commission proceedings. |
| Australia announces proposed frontier AI safety legislation | Government of Australia | 2026 | Government proposal — not enacted | The government announced proposed mandatory frontier AI safety and security standards and a targeted AI safety regulator. Proposed measures include risk management, independent evaluation and incident notification. As of 15 November these are announced proposals, not enacted obligations. |
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
2026-09-30
OAIC automated-decision transparency resources
Regulator guidance supporting already enacted requirements, not a new Act. The Privacy and Other Legislation Amendment Act 2024 enacted APP privacy-policy transparency requirements commencing 2026-12-10. They apply where an APP entity arranges for a computer program to make, or do something substantially and directly related to making, a decision that could reasonably significantly affect an individual’s rights or interests, and uses personal information about that individual in the program. As of 2026-10-03 these requirements are enacted but not yet applicable.
2026-10-07
OAIC opens investigation into HeyCyan smart-glasses app maker
OAIC commenced a privacy investigation into Shenzhen Qingcheng Future Technology, maker of the HeyCyan app, after unanswered preliminary inquiries. Investigation initiation is an enforcement/watchlist signal under existing privacy law, not an established violation, imposed penalty or new AI statute.
2026-10-07
DTA reports inaugural AI Review Committee advice
DTA’s 7 October 2026 announcement reports the AI Review Committee’s inaugural meeting held on 26 June. The committee gives expert, non-binding advice on selected public-sector AI uses before deployment and reviews agencies’ responses to serious AI incidents. Agencies retain implementation, decision-making and risk accountability; this is governance implementation, not new legislation.
2026-10-20
Fair Work Commission requirements for GenAI in cases
Participants using GenAI to prepare Commission case documents must disclose its use and check accuracy and relevance; witness statements must reflect personal knowledge and truthful words. The requirements concern Commission proceedings.
2026-10-30
Australian Government AI Reference Architecture
DTA announced a reference architecture for public agencies designing and governing AI systems, including assurance and human oversight. The architecture is guidance, not a national AI statute.
2026-11-10
Australia announces proposed frontier AI safety legislation
The government announced proposed mandatory frontier AI safety and security standards and a targeted AI safety regulator. Proposed measures include risk management, independent evaluation and incident notification. As of 15 November these are announced proposals, not enacted obligations.
2026-11-10
Government releases AI incident Rapid Review recommendations
The government released 18 Rapid Review recommendations following an AI-agent incident affecting an Australian government health website. The recommendations are implementation activity, not a finding that a new AI statute has commenced.
2026-11-12
APRA identifies AI-related operational risks
APRA identifies AI and material service-provider dependencies as changing operational risk and is intensifying supervisory engagement. The outlook is a supervisory assessment, not a new AI rule.
计划中的里程碑
企业影响
企业不应将封闭式强制性保障措施咨询视为现行法律。近期管控措施应侧重于现有的隐私、消费者、行业和安全义务;在2026年12月10日之前,确保重大自动化决策的透明度和记录保存;模型和数据文档;人工审核和升级机制;测试和保障;采购和供应商管控;以及监测未来联邦或行业特定规则的出台。
OAIC’s 30 September resources translate the future statutory requirement into operational guidance. APP entities using personal information in decisions that could reasonably significantly affect rights or interests should inventory relevant processes and prepare privacy-policy disclosures ahead of 10 December 2026. The guidance did not enact a new law.
Public agencies can use the reference architecture for design and assurance; participants in Fair Work Commission cases should follow its case-document requirements.
Frontier AI developers should monitor consultation and legislative detail on proposed safety testing, risk management and incident notification. Financial institutions should monitor APRA supervisory engagement on AI-related operational risks.
观察站解读
澳大利亚是一个以实施为主导且适应性强的司法管辖区。目前,其在人工智能领域的约束性信号主要体现在联邦公共部门政策和有针对性的隐私透明度方面,而国家人工智能计划和澳大利亚人工智能安全研究所则致力于提升能力和指明方向。缺乏广泛实施的高风险防护机制本身就是一个重要问题,应作为一项政策现状予以记录,而不应被悄然视为监管漏洞或假定的未来法律。
官方资源
| Resource | 来源 | Type | 日期 | Legal force | 为何重要 |
|---|---|---|---|---|---|
国家人工智能战略及实施方向。 | Australian Government / Department of Industry, Science and Resources | Strategy | 2025-12-02 | Not applicable | 为澳大利亚在人工智能能力、安全、应用和劳动力发展方面制定战略方向,而不创设独立的约束性义务。 |
拟议的强制性高风险人工智能防护措施的正式咨询和政策状态记录。 | Australian Government / Department of Industry, Science and Resources | Guidance | 2024–2026 | Not applicable | 必须将其视为已关闭/已延期的提案途径,而不是已颁布的法规。 |
联邦政府公共部门强制性人工智能政策已纳入范围。 | Australian Government / Digital Transformation Agency | Public-sector rule | 2025 | Binding | 为澳大利亚政府人工智能的使用提供了目前最清晰的强制性操作控制措施。 |
官方人工智能安全与评估机构页面。 | Australian Government / Department of Industry, Science and Resources | Framework | 2025–2026 | Not applicable | 跟踪国家安全测试和评估能力,但不暗示一般监管权力。 |
Official APP 1 guidance explains the enacted 2026-12-10 automated-decision transparency requirements. The guidance is not itself legislation. | Office of the Australian Information Commissioner | Guidance | 2025–2026 | Guidance | Read as regulator implementation support for an enacted future statutory duty, not as a new Act. |
监管机构咨询支持实施自动化决策透明度要求。 | Office of the Australian Information Commissioner | Guidance | 2025–2026 | Guidance | 信号表明可能采取的措施,但其本身并不构成具有约束力的法规。 |
Regulator guidance supporting already enacted requirements, not a new Act. The Privacy and Other Legislation Amendment Act 2024 enacted APP privacy-policy transparency requirements commencing 2026-12-10. They apply where an APP entity arranges for a computer program to make, or do something substantially and directly related to making, a decision that could reasonably significantly affect an individual’s rights or interests, and uses personal information about that individual in the program. As of 2026-10-03 these requirements are enacted but not yet applicable. | Office of the Australian Information Commissioner | Guidance | 2026-09-30 | Guidance | Inventory relevant automated-decision processes and prepare APP privacy-policy disclosures before commencement. |
The Privacy and Other Legislation Amendment Act 2024 enacted APP privacy-policy transparency requirements commencing 2026-12-10. They apply where an APP entity arranges for a computer program to make, or do something substantially and directly related to making, a decision that could reasonably significantly affect an individual’s rights or interests, and uses personal information about that individual in the program. As of 2026-10-03 these requirements are enacted but not yet applicable. | Parliament of Australia | Law | 2024-12-10 | Binding | Primary statutory source; Part 15 transparency duties are enacted but apply from 2026-12-10. |
Official notice initiating a privacy investigation. The announcement does not establish a violation or penalty and does not enact a new AI law. | Office of the Australian Information Commissioner | Enforcement case | 2026-10-07 | Not applicable | Tracks privacy-enforcement scrutiny of AI-enabled wearables while reserving conclusions until official findings. |
DTA’s 7 October 2026 announcement reports the AI Review Committee’s inaugural meeting held on 26 June. The committee gives expert, non-binding advice on selected public-sector AI uses before deployment and reviews agencies’ responses to serious AI incidents. Agencies retain implementation, decision-making and risk accountability; this is governance implementation, not new legislation. | Digital Transformation Agency | 报告 | 2026-10-07 | Not applicable | Backfills advisory oversight and incident-learning capacity without changing statutory classifications. |
Defines non-binding pre-deployment advice on selected agency AI uses and review of serious-incident responses; agencies remain accountable. | Digital Transformation Agency | Framework | Guidance | Read alongside applicable Commonwealth policy, not as an independent AI statute. | |
DTA announced a reference architecture for public agencies designing and governing AI systems, including assurance and human oversight. The architecture is guidance, not a national AI statute. | Digital Transformation Agency | Framework | 2026-10-30 | Guidance | Non-mandatory public-sector reference architecture |
Participants using GenAI to prepare Commission case documents must disclose its use and check accuracy and relevance; witness statements must reflect personal knowledge and truthful words. The requirements concern Commission proceedings. | Fair Work Commission | Guidance | 2026-10-20 | Binding | Case-procedure requirements |
The government announced proposed mandatory frontier AI safety and security standards and a targeted AI safety regulator. Proposed measures include risk management, independent evaluation and incident notification. As of 15 November these are announced proposals, not enacted obligations. | Government of Australia | Guidance | 2026-11-10 | Proposed | Official evidence for the stated development and its legal status. |
The government released 18 Rapid Review recommendations following an AI-agent incident affecting an Australian government health website. The recommendations are implementation activity, not a finding that a new AI statute has commenced. | Government of Australia | 报告 | 2026-11-10 | Not applicable | Official evidence for the stated development and its legal status. |
APRA identifies AI and material service-provider dependencies as changing operational risk and is intensifying supervisory engagement. The outlook is a supervisory assessment, not a new AI rule. | Australian Prudential Regulation Authority | 报告 | 2026-11-12 | Not applicable | Official evidence for the stated development and its legal status. |