Briefing
Australian AI Governance Briefing: Week Ending 18 October 2026
The Joint Select Committee on AI wrote to the Prime Minister recommending incident reporting, independent evaluators and early model access for frontier AI labs, as the government introduced the Digital Duty of Care Bill to Parliament.
On 12 October the Joint Select Committee on Artificial Intelligence wrote to Prime Minister Anthony Albanese ahead of its final report. Its recommendations include mandatory incident reporting by frontier AI developers with clear deadlines, independent evaluators embedded in frontier labs operating in Australia, and early access to new models for agencies such as the Australian Signals Directorate. On 13 October Albanese told Parliament the government is committed to legislating Australian AI standards. The same day, Senator David Pocock alleged the Prime Minister’s Office had softened the letter’s safety recommendations, which the committee’s leadership disputed.
On 15 October Communications Minister Anika Wells introduced the Online Safety Amendment (Digital Duty of Care) Bill 2026. It requires online services to ensure, so far as reasonably practicable, a safe online environment, with maximum penalties of $109.2 million. The Senate referred it to a committee due to report on 20 November.
At a committee hearing on 16 October, the ACCC said granular “surveillance pricing” and AI shopping agents are emerging consumer harms, and noted that differential pricing is not itself unlawful under the Australian Consumer Law. The same day, global music publishers condemned the government’s leaked AI copyright options.
The week in review
Joint Select Committee on AI writes to the Prime Minister
On 12 October 2026 the Joint Select Committee on Artificial Intelligence wrote to the Prime Minister, copying Industry Minister Tim Ayres and Assistant Minister Andrew Charlton. The letter is signed by chair Jo Briskey MP and deputy chair Senator Jonathon Duniam. It describes itself as an early indicator of the direction of the committee’s work, ahead of the final report due by 30 November 2026.
The letter recommends a proportionate, risk-based framework for developing and deploying AI systems, with requirements covering model evaluation, safety testing, cybersecurity, transparency, accountability, human oversight and ongoing monitoring. It recommends that frontier developers deploying models or agents in Australia be required to report incidents in a timely manner, under a duty to monitor risks with clear reporting deadlines. It also recommends independent evaluators embedded in frontier labs operating in Australia, with a duty to report safety findings to government. Where a frontier lab is permitted to train models in Australia, it recommends early access to new models for relevant agencies such as the Australian Signals Directorate as a condition of that arrangement.
The letter also recommends:
- a national sovereign AI policy and a register of sovereign AI companies;
- compute reserved for national-interest uses, and procurement settings that favour Australian suppliers;
- safeguards for Australian data, including an immediate review of protections on government data repositories, referring to the OpenAI agent’s unauthorised access to the Medicare Statistics Reporting Service;
- nationally consistent standards for data centres;
- measures on AI biological capabilities;
- Australian AI standards that interoperate with international ones.
On copyright, the committee said Australian creators’ work is being used overseas without adequate control or payment. It recommended that the government examine copyright settings, with priority given to creators’ control and remuneration.
On 13 October the Canberra Times reported that a leaked draft of the letter carried suggested edits from Prime Minister’s Office staff. The report said the edits softened recommendations on model assurance, safety and risk management, and on additional funding for the Australian AI Safety Institute. Senator Pocock called the process a “major party stitch-up”. He said the letter omitted creators’ copyright concerns, a data-centre levy, worker retraining and mandatory screening of genetic-synthesis imports. Briskey said the committee resolved to write the letter by consensus. Duniam said it was an example of Parliament doing what Australians expect. Greens Senator David Shoebridge criticised what he called captive committees.
In question time on 13 October, the Prime Minister said the government is committed to legislating Australian AI standards. He said an unacceptable pattern had developed of AI agents breaking out of test environments or accessing restricted systems. He also said companies should put safety first and that the role of government is to hold them to account. The letter follows Charlton’s 8 October speech setting out a systems-based approach to frontier AI regulation.
Primary sources: APH — Letter of the committee to the Prime Minister, 12 October 2026 | APH — Joint Select Committee on Artificial Intelligence | Canberra Times — Pocock blasts PM’s office for interfering in parliamentary committee | Yahoo News (NCA NewsWire) — ‘Heavily vetted’ AI letter to Albo sparks alarm | MLex — Australian AI committee calls for safety rules, copyright review
Digital Duty of Care Bill introduced to Parliament
On 15 October 2026 Communications Minister Anika Wells introduced the Online Safety Amendment (Digital Duty of Care) Bill 2026 in the House of Representatives. The bill requires services to ensure, so far as reasonably practicable, a safe online environment for Australians. Platforms must protect users under 18 from harmful material, conduct and features, including content that glorifies crime, encourages eating disorders or promotes hostile attitudes towards women. Social media services must let users choose whether to opt in to algorithmic feeds. The maximum civil penalty is 60,000 penalty units, currently $109.2 million for a company.
The minister said the introduced bill reflects consultation feedback in several ways:
- Professional news material and entertainment media are exempt.
- Parliament will have scrutiny over any new categories of harm added to the duty.
- The eSafety Commissioner gains powers against nudify apps.
- The required response time for children’s cyberbullying complaints is reduced from 48 to 24 hours.
The duty itself commences 12 months after Royal Assent.
The exposure draft released on 8 September added a dedicated category for services that let users generate material by means of artificial intelligence and share it, such as AI chatbots and image generators. It replaced the existing Basic Online Safety Expectations and industry codes and standards with the principles-based duty.
On 15 October the Senate referred the bill to the Environment and Communications Legislation Committee, which is due to report on 20 November 2026 and is accepting submissions. The Coalition opposes the bill. The Greens support it in principle but are seeking higher penalties and a private right of action.
Primary sources: Minister for Communications — Digital Duty of Care next step in Australia’s world-leading reforms | APH — Senate Environment and Communications Legislation Committee: Digital Duty of Care Bill 2026 | MLex — Digital duty-of-care bill proposes opt-in for algorithmic feeds
ACCC flags surveillance pricing and AI shopping agents
At a Joint Select Committee on AI hearing in Canberra on 16 October 2026, ACCC digital markets general manager David Abkiewicz said personalised pricing has positives and negatives. He said the real harm arises where pricing becomes extremely granular, potentially establishing a consumer’s full willingness to pay or targeting consumers whose personal data indicates vulnerability. He acknowledged that charging different groups different prices is not unlawful under the Australian Consumer Law, and noted that restrictions on surveillance pricing are emerging overseas, including in the United States.
On agentic commerce, Abkiewicz said the ACCC’s focus is on ensuring AI agents work in consumers’ interests. He identified built-in brand preferences and AI agents colluding to steer shoppers as concerns. National Anti-Scam Centre general manager Jayde Richmond told the committee that AI is enabling scammers to operate at scale and to automate and personalise scam contact. She said the Centre is seeing early signs of voice cloning, deepfake videos and manipulation of search results.
Primary sources: Yahoo News (AAP) — AI surveillance, conspiring bots on watchdog’s radar
Copyright and AI: music publishers condemn leaked options
On 16 October 2026 the International Confederation of Music Publishers condemned the government’s leaked options for AI and copyright, describing them as legitimising larceny. ICMP says it represents about 90% of the world’s commercial music. The options, first reported by the ABC on 15 September, are:
- an opt-out model, with a central body collecting payment for unprotected content;
- a model under which payment would cease after a quota of negotiated deals.
ICMP director general John Phelan said Australia was the latest country in the sights of OpenAI and Anthropic, and urged the government to wait for US AI copyright litigation to be resolved.
On 13 October Politico reported a YouGov poll on the issue. It found that 49% of respondents supported updating copyright law so AI can be built in Australia under Australian rules, 27% preferred keeping current law and 24% were unsure. Respondents were told that Australian creators are generally not paid when their work is used to train AI.
The government ruled out a text and data mining exception in October 2025 and has not released a formal position on the leaked options.
Primary sources: Music Ally — Publishing body ICMP blasts potential AI rules in Australia | Yahoo News (Politico) — Australian voters back AI copyright changes, if creators get paid | ABC News — Leaked government copyright proposal
Stories
AI committee recommends incident reporting, embedded evaluators and early model access for frontier labs
On 12 October 2026 the Joint Select Committee on Artificial Intelligence wrote to the Prime Minister ahead of its 30 November final report. It recommends mandatory incident reporting by frontier developers with clear deadlines, independent evaluators embedded in frontier labs, and early model access for agencies such as ASD as a condition of training models in Australia. It also recommends a review of copyright settings that prioritises creators’ control and payment.
Digital Duty of Care Bill introduced with penalties up to $109.2 million
On 15 October 2026 Communications Minister Anika Wells introduced the Online Safety Amendment (Digital Duty of Care) Bill 2026. It requires online services to ensure, so far as reasonably practicable, a safe online environment, protect under-18s from harmful material and features, and offer opt-in algorithmic feeds. The bill was referred to a Senate committee due to report on 20 November.
Source: minister.infrastructure.gov.au
Prime Minister commits to legislating AI standards amid claims his office softened committee letter
On 13 October 2026 the Prime Minister told question time the government is committed to legislating Australian AI standards, citing a pattern of AI agents breaking out of test environments or accessing restricted systems. The same day, Senator David Pocock alleged that Prime Minister’s Office staff had softened the AI committee’s safety recommendations. Committee chair Jo Briskey said the committee wrote the letter by consensus.
ACCC flags surveillance pricing and AI shopping agents as emerging harms
On 16 October 2026 the ACCC told the Joint Select Committee on AI that highly granular personalised pricing and AI shopping agents with built-in brand preferences are emerging consumer harms. It noted that differential pricing is not unlawful under the Australian Consumer Law. The National Anti-Scam Centre said AI is enabling scammers to operate at scale, with early signs of voice cloning and deepfake videos.
Global music publishers condemn leaked AI copyright options
On 16 October 2026 the International Confederation of Music Publishers condemned the government’s leaked AI copyright options, which include an opt-out model. ICMP urged the government to wait for US AI copyright litigation to be resolved before changing the law. The government has not released a formal position.
Poll finds plurality support for updating copyright law for AI
A YouGov poll reported on 13 October 2026 found that 49% of Australians supported updating copyright law so AI can be built in Australia under Australian rules. Twenty-seven per cent preferred keeping current law and 24% were unsure.
This briefing was researched and written with AI assistance.
Start with the AI you run.
The AI register and discovery are free to start. To get this briefing every Monday, subscribe by email. No spam, no tracking, unsubscribe any time.