All stories
The front page and the section pages show what is current. This page lists everything we have published, so a story stays findable after it stops being news.
Research
Answers to the Joint Select Committee on AI, published in November, show what has not been measured: the Productivity Commission has not modelled AI job losses or payments to overseas AI providers, and the workplace department holds no data on whether employers consult workers before bringing in AI.
Data
The 139,000 in the Jobs and Skills Report 2026 is a modelled scenario: fewer Skill Level 1 workers than baseline growth if education choices followed labour demand, offset by about 106,000 more trade and technical workers. JSA says the scenario does not consider AI, and the report says its baseline does not explicitly model it.
Government
A communique from the Secretaries Digital and Data Committee, published on 6 November, cites the risk of relying on a small number of overseas AI providers and says open-weight models hosted on government-controlled infrastructure can be a secure, cost-effective option for routine work. A week earlier Home Affairs told a parliamentary committee its broader posture on open-weight models is still being worked through.
Government
The DTA published its AI Reference Architecture on 30 October. Its companion guide works through Microsoft 365 Copilot and tells agencies an Australian data centre commitment for data at rest does not necessarily mean all Copilot inference happens in Australia. Microsoft's own page says customers outside the EU may have queries processed in the US, EU or other regions.
Regulation
The OAIC received 5,268 privacy complaints in 2025-26, up 60 per cent. Its annual report says the rise appears to be attributable to concern about privacy, especially about social media, AI companies and data brokers, citing its attitudes survey. Its table of complaints by sector has no line for AI.
Government
Version 2.1 of the Policy for the responsible use of AI in government, published 21 October 2026, changes no requirement. It adds Appendix D, a classification of AI use that agencies must apply in their registers, impact assessments and transparency statements: seven usage patterns where there were four, including coding, workflow automation and strategic support, and a decision-making category redefined around effect on rights or interests. The DTA's impact assessment tool still offers the old four boxes.
Regulation
Who is responsible under Australian privacy law when smart glasses record you? The Privacy Commissioner says it is the entity holding the personal information, likely the software provider, and a retailer or manufacturer that collects nothing may have no Privacy Act obligations. Most collection needs no consent. On 7 October the OAIC announced an investigation into the maker of the HeyCyan app after it did not answer preliminary inquiries.
Policy
The Online Safety Amendment (Digital Duty of Care) Bill 2026 was introduced on 15 October and sent to a Senate committee, whose submissions closed on 30 October. Against the September draft, a service that generates material using AI is now an online service in its own right and pre-deployment testing appears in the explanatory memorandum. Schedule 3 still repeals every industry code and standard, with no saving item.
Regulation
In a 12 October letter to the Prime Minister, the Joint Select Committee on Artificial Intelligence recommends mandated independent evaluators inside frontier labs operating in Australia, incident reporting with clear deadlines, early access to new models for agencies such as ASD, a standing parliamentary committee, and pro-Australian procurement rules. Its full report is due by 30 November.
Research
A CSIRO report the AI Safety Institute published on 8 October finds an AI can be right on the task it is checked on and still shift its overseer's beliefs about something else. Its experiments used AI verifiers, and the clear steering result came when an adversary controlled the verifier's training. The Institute's summary page leaves that condition out.
Regulation
ASIC's 30 September letter to bank boards lists a banking sector AI review among its new activities for 2026-27, focused on new and proposed use cases of AI by banks and the impact on customers. It is the only one of ASIC's five sector letters to give AI its own heading and an expected start. The general insurance letter does not mention AI.
Regulation
ASIC's amended market integrity rules, registered 17 September 2026, bar a market participant from acting on a client's order it ought reasonably suspect will have the effect of creating a false or misleading appearance. The consultation paper proposed an intention test only; ASIC added the effect limb afterwards, citing agentic AI. The rules commence on 18 March 2028.
Government
The DTA points to its agentic AI addendum as the foundation for Australia's part in a new OECD working group. Read in full, 14 of its 25 new criteria sit under 'Agencies must': a human accountable for agents' decisions, oversight that is human-in-the-loop or human-on-the-loop, kill switches. It calls itself best practice guidance, and the binding AI policy lists the standard it extends under recommended actions.
Research
eSafety surveyed 1,950 children aged 10 to 17 before the Age-Restricted Material Codes for AI services took effect. 78% had used an AI assistant and 8% an AI companion; among assistant users, 46% had sought personal advice. One in five who had used either had a potentially inappropriate or harmful interaction.
Government
The AI Review Committee's first meeting, on 26 June, considered two staff-facing AI uses at the Attorney-General's Department and Services Australia. The DTA announced it on 7 October, 103 days later, with terms of reference under which the Minister for Finance receives the minutes and the advice issued to agencies, and the public gets communiques. The communique says it is not an endorsement or approval.
Policy
The exposure draft Privacy Amendment (Personal Data Protection) Bill 2026 never uses the words artificial intelligence. Its AI rule is a definition: personal information is collected however it is obtained, including by generating or deriving it, which the consultation paper says will make AI inferences count as collection. That puts them under a new fair and reasonable test and, for sensitive inferences, a consent rule with exceptions. It is a draft; submissions closed 18 September.
Policy
The exposure draft Online Safety Amendment (Digital Duty of Care) Bill 2026, released 8 September, repeals the Online Safety Act's industry codes and standards and its basic online safety expectations, the instruments that carry Australia's generative AI safety duties, with no provision keeping them in force. In their place is one general duty, backed by 60,000 penalty units. The bill had not reached Parliament as at 6 October.
Government
The NSW Auditor-General's Internal controls and governance 2026 report examined 10 agencies with the highest AI use or potential benefit. All 10 have addressed accountability for AI and eight train staff, but half have an AI policy, four have considered AI in IT policies and one in procurement. The rules changed just before the report was published: a mandatory AI Operational Policy now requires registration and an annual attestation, the first due 31 October 2027.
Policy
The OAIC's 30 September guidance on the APP 1.7 transparency obligation: rule-based software counts, human review does not take a decision out of scope, and the in-doubt instruction is to disclose.
Policy
PSPF Direction 002-2026, published 29 September 2026, names AI in its title and frontier AI in its reasoning, then requires every non-corporate Commonwealth entity to stocktake its legacy technology and send Home Affairs a risk management plan with a reduction target by 31 March 2027. Systems of Government Significance report by 31 December 2026. No AI tool is banned, approved or named.
Policy
PSPF Policy Advisory 001-2026, Cyber Security Readiness in the Frontier AI Era, says Australian Government entities do not need access to the most advanced frontier AI models to stay protected. It describes a vulnerability storm and exploit timelines collapsing from days to hours, and answers with the Essential Eight, the ISM and existing PSPF requirements. In ASD's six-strategy annex, adopting AI for cyber defence is fifth and medium-term.
Policy
The Senate Select Committee on Adopting AI tabled 13 recommendations on 26 November 2024. Senate practice requires a government response within three months. It was tabled in the House of Representatives on 1 April 2026, sixteen months later, in five thematic parts that never address the recommendations by number. The Senate's own status pages still do not record it.
Policy
PSPF Direction 001-2025 bans DeepSeek products, applications and web services from Australian Government systems, but a footnote excludes open source LLMs meeting three conditions. Since Direction 004-2025 took effect on 31 October 2025 the banned list has been the Commonwealth Technology Standard Deny List, which is not in the PSPF publications library.
Policy
PSPF Policy Advisory 001-2025 confirms OFFICIAL information can be used with generative AI, and names OpenAI and Anthropic alongside Hosting Certification Framework providers as not requiring an additional Foreign Ownership, Control or Influence assessment. All other providers must be assessed under PSPF Direction 001-2024 before staff get access.
Policy
Two frameworks bind Commonwealth AI use, not one. The Policy for the responsible use of AI in government reached version 2.0 on 15 December 2025 and carries eight mandatory requirements for non-corporate Commonwealth entities. The Protective Security Policy Framework binds the same entities, and its Directions include a standing ban on DeepSeek products and a requirement to adhere to the Commonwealth Technology Standard. The APS AI Plan, the AI technical standard, the agentic addendum and the PSPF policy advisories are guidance.
Regulation
ASIC tells licensees that cyber risk management must be demonstrably effective and proportionate to the size, nature and complexity of a business, as set out in the FIIG judgment. We read the judgment. Those words are not in it. What is in it is a five-factor test that is more concrete and more useful, plus an explicit statement that the law does not require perfection.
Government
Queensland's Auditor-General found the department responsible for AI policy has limited visibility of AI use across government. Five weeks later NSW's auditor counted 357 AI tools across 21 agencies. Different methods, same blind spot.
Policy
The automated decision-making transparency obligation commences 10 December 2026 under the Privacy Act. It has a three-limb test, and the limb that decides scope is whether a decision could reasonably be expected to significantly affect a person's rights or interests. The OAIC ran an Issues Paper consultation on what that means; it closed 15 June 2026 and the guidance is not out.
Policy
On 10 August the Australian AI Safety Institute published its first publication, 119 pages by Gradient Institute on risks and controls for multi-agent systems. It answers a question we left open on 31 July, and its hardest conclusion is that the controls for AI agents meeting across organisations depend on infrastructure no single organisation can build. There is no Let's Encrypt for agent identity, and the report says nobody is positioned to become one.
Policy
Australia's AI memorandum with Singapore was signed in December 2024, well before the Canada and UK ones. On 30 July 2026 it produced the first Australia-Singapore Cyber and Digital Senior Officials Dialogue. We read the memorandum in full: it creates no binding obligations, commits no money, and its first review window opens on 16 December 2026.
Policy
We read both Directions to Comply on the eSafety register. Each targets a high impact generative AI DIS with a tier 1 risk profile, each cites the same two compliance measures, and each is about age assurance on AI-generated pornography.
Policy
Schedule 6 of the Age-Restricted Material codes has bound designated internet services since 9 March 2026. It names AI companion chatbots, deems some generative services highest-risk without an assessment, and creates a regulated category for platforms that host uploaded models.
Policy
The Online Safety (Designated Internet Services) Industry Standard 2024 defines a 'high impact generative AI DIS' and binds it: prevent prohibited outputs, test models, differentiate AI outputs, and assess before a material change. Registered June 2024, in force 22 December 2024.
Policy
The Relevant Electronic Services industry standard, covering messaging and email, was determined by the same Commissioner on the same day as the standard that regulates generative AI. It does not use the word generative once. What it does say about AI, and what that means.
Policy
On 25 May 2026 Australia and the UK signed a memorandum of understanding on AI safety and security. The department published a three-bullet overview, not the instrument. Eleven weeks earlier it published the Canada MoU in full, so we know what these agreements usually contain and what the overview leaves out.
Policy
The ACITI Partnership was announced in November 2025 with an emphasis on green energy innovation, resilient supply chains and critical minerals, and AI as something it would also examine. The memorandum of understanding whose signing was announced in July 2026 sets up four working streams, all of them AI, and does not mention minerals, energy or net zero at all.
Policy
The Australia-Canada memorandum of understanding on AI safety is published in full. It creates no legally binding obligations, has no financial implication for either country, and can be ended on six months' notice. It also sets two hard deadlines, and the second one falls around 1 September 2026.
Policy
Australia's AI Safety Institute was announced on 25 November 2025 and sits inside the department it advises. Its technical partners include the Australian Signals Directorate. The multi-agent risk project cited as its work was published four months before it existed and funded by the department. And the international network it belongs to has dropped the words AI Safety from its name.
Policy
On 20 July 2026 six ministers jointly named the Albanese Government's AI safety priorities: a Digital Duty of Care, a second tranche of privacy reform, workplace AI safety, consumer law options, and a framework for automated decision-making in federal agencies. Four lead ministers share them, the release gave none a date, and two have since reached draft bills.
Policy
The OAIC has updated its guidance on facial recognition in retail spaces to implement the Administrative Review Tribunal's findings in the Bunnings matter, which covered 62 stores between 2018 and 2021. The Privacy Commissioner says a precautionary approach is required and that each deployment must still be assessed individually. The Kmart determination remains under review with hearings in early 2027.
Research
The OAIC's triennial Australian Community Attitudes to Privacy Survey put AI companies equal-last on trust at 4 per cent. Ninety-six per cent want conditions before AI makes a decision about them, and every condition measured in both years rose. We read the 2026 report and set it against the policy Australia actually chose.
Data
The National AI Centre and CSIRO's census of Australia's AI ecosystem counts 1,533 companies, 25 urban clusters led by Melbourne CBD, patents nearly quadrupling, and a commercialisation gap of almost 23 publications per patent. The numbers, charted from the report.
Regulation
ASIC Commissioner Simone Constant's 8 May 2026 open letter tells every AFS licensee that frontier AI is accelerating cyber attacks now, lists twelve expected actions, and anchors the lot to the FIIG Securities judgment. What a board must be able to show.
Data
DEWR's new monitoring report is the first Australian government data on AI and employment: exposed occupations grew 5.6 per cent against 9.5 per cent for the least exposed, the effect is not robust, software jobs are up 25 per cent, and the most exposed workers are majority women and degree-holders.
Safety
OpenAI's system card rates all three GPT-5.6 models High capability for cybersecurity and biological risk. Washington reportedly held the launch for testing. In Australia the same models arrived under no AI-specific check, which is the government's explicit design.
Policy
In September 2024 the government proposed ten mandatory guardrails for high-risk AI and three ways to legislate them, including an Australian AI Act. The National AI Plan commits to none of them, and the word guardrails does not appear once in its 37 pages.
Government
The NSW Auditor-General's 2025 internal controls report found 21 of the state's 26 largest agencies using 357 AI tools, while only 38 per cent have a formal AI policy and 5 per cent have AI procurement guidance. The numbers, from the audit itself.
Business
OpenAI's own documentation lists Australia as a data residency region with storage yes, processing no: prompts are still processed on offshore infrastructure, and residency endpoints carry a 10 per cent price uplift for new models. What that means before you tick the compliance box.
Industry
WiseTech Global's 1H26 ASX release announces a phased headcount reduction of up to 50 per cent in product, development and customer service, about 2,000 roles, framed as becoming an AI-led organisation. What the filing actually says, and the caveats it carries.
