AI Australia Watch

Campaign wins

Campaign Success Stories

These case studies show how Australian communities, advocacy organisations, legal services and regulators have successfully challenged AI-related infrastructure, automated systems, facial recognition and policy proposals.

Hazelmere Data Centre, Western Australia

GreenSquareDC withdrew the development application on 14 May 2026, before the scheduled Development Assessment Panel hearing.

What was proposed

GreenSquareDC proposed a 120 MW data centre in Hazelmere, near the Mandoon Bilya / Helena River, the Helena River Waldorf School, Trillion Trees Australia's community nursery, Aboriginal heritage sites and a conservation wetland.

Community concerns included noise from backup diesel generators, impacts on the river and surrounding environment, proximity to the school and homes, cultural heritage, and the suitability of the site.

What people did

  1. Built a local campaign around the development application. Trillion Trees Australia established the Save the Mandoon Bilya campaign and published information about the proposal, planning process and ways people could participate.
  2. Formed a coalition. Trillion Trees worked alongside the Bibbul Ngarma Aboriginal Association, Guildford Association, Midland Society, residents and other local groups.
  3. Created a public petition. Around 4,700 people signed the campaign petition.
  4. Turned petition support into formal planning objections. The City of Swan received almost 1,900 submissions, with more than 1,800 objections recorded.
  5. Held a public rally. Around 120 people attended a rally outside the City of Swan administration building and marched to Midland Town Hall.
  6. Put technical and local evidence into the campaign. Residents and campaigners raised specific issues including generator noise, environmental impacts, proximity to the river and school, Aboriginal heritage and planning suitability.
  7. Engaged elected representatives and council. The issue was taken through the formal council and Development Assessment Panel process. City of Swan councillors voted 11-2 for the mayor to make a deputation opposing the proposal.
  8. Kept pressure on the formal decision-making process. City of Swan planning staff ultimately recommended that the Development Assessment Panel refuse the application.

Outcome

GreenSquareDC withdrew the development application on 14 May 2026, before the scheduled Development Assessment Panel hearing.

The project therefore did not proceed under the submitted application.

Sources

North Katoomba Data Centre, New South Wales

The applicant withdrew Development Application X/419/2026 in June 2026.

What was proposed

A data centre was proposed for 41-45 Barton Street in North Katoomba, close to homes, Katoomba North Public School, Yosemite Creek and surrounding bushland.

Residents raised concerns about low-frequency noise, the height of the proposed development, generator and battery infrastructure, bushfire risk, environmental impacts and compatibility with the surrounding area.

What people did

  1. Raised the issue locally as soon as residents became aware of the application. A local resident posted information about the proposal in a community Facebook group and created a dedicated page so concerned residents could organise.
  2. Created community campaign groups. Stop Data Centres Blue Mountains and No Data Centres in Katoomba became focal points for organising and sharing information.
  3. Door-knocked nearby residents. Campaigners spoke directly with people living around the site, including residents who had not known the development application had been lodged.
  4. Distributed printed flyers. This ensured residents who were not active online still received information about the proposal and how to respond.
  5. Created online action groups and petitions. The main campaign petition gathered more than 730 signatures, while reporting across the campaign described close to 1,000 petition signatures.
  6. Held community meetings. Residents met to discuss the proposal, planning documents and possible responses.
  7. Read and analysed the development application documents. Campaigners examined technical material including acoustic, geotechnical and planning documents rather than relying only on general objections.
  8. Gave residents specific instructions for formal objections. The campaign published the development application number, council email address, submission deadline and a list of planning issues residents could address.
  9. Lodged formal objections with Blue Mountains City Council. Dozens of objections were submitted through the council process.
  10. Brought local organisations into the campaign. The Katoomba North Public School P&C voted unanimously to oppose the proposal.
  11. Engaged councillors and political representatives. Councillors publicly questioned the proposal and Mayor Mark Greenhill criticised the suitability of the location.

Outcome

The applicant withdrew Development Application X/419/2026 in June 2026.

The developer stated that it had listened to views expressed by residents, community groups and other stakeholders and chose to step back from the application.

Sources

Australian Creators and the Proposed AI Text and Data Mining Copyright Exception

On 26 October 2025, the Australian Government formally announced that it was not considering a text and data mining exception to Australian copyright law.

What was proposed

In 2025, the Productivity Commission considered whether Australia should introduce a text and data mining copyright exception.

Such an exception could have allowed copyright-protected material to be used for activities including AI training without the ordinary requirement to obtain permission from individual rights holders.

What people did

  1. Creative-sector organisations responded publicly as soon as the proposal emerged. Groups representing authors, illustrators, musicians, publishers and other creators published detailed objections.
  2. Made formal submissions to the Productivity Commission. The Australian Society of Authors, APRA AMCOS and other organisations submitted evidence explaining the effects the proposed exception could have on copyright holders and existing licensing markets.
  3. Mobilised individual creators. Industry organisations briefed members on the proposal and provided information about how to participate in the consultation.
  4. Appeared before parliamentary inquiries. Authors, musicians, industry representatives and copyright experts gave evidence at Senate hearings.
  5. Used recognisable Australian creators as public witnesses. Participants included authors Anna Funder, Thomas Keneally, Caroline Overington and Sally Rippin, alongside industry representatives and AI academic Professor Toby Walsh.
  6. Coordinated across different creative industries. Authors, publishers, songwriters, composers, musicians and Indigenous music organisations pushed the issue through their respective peak bodies.
  7. Proposed an alternative mechanism. Rather than only objecting to the exception, organisations argued that AI developers should use licensing arrangements to access copyright-protected work.
  8. Maintained pressure through the formal consultation period. Advocacy continued through submissions, hearings, media coverage and engagement with government.

Outcome

On 26 October 2025, the Australian Government formally announced that it was not considering a text and data mining exception to Australian copyright law.

The Productivity Commission's final report subsequently recommended monitoring AI licensing markets rather than introducing the proposed exception at that time.

This was a specific policy win against the broad TDM exception proposed in 2025. Separate AI copyright reform options are again being debated in 2026, so the broader copyright issue remains active.

October 2026 update: APRA AMCOS, ARIA and PPCA have responded to the parliamentary AI committee’s copyright proposals, renewing calls for creator consent, licensing and payment. APRA AMCOS response · ARIA and PPCA response.

Sources

CHOICE Campaign on Facial Recognition in Australian Retail Stores

The Good Guys paused its facial recognition trial in June 2022. Bunnings and Kmart also paused their systems amid the public controversy.

What was happening

CHOICE investigated the use of facial recognition technology by major Australian retailers and found that Bunnings, Kmart and The Good Guys were using systems that captured customers' biometric information in stores.

The investigation raised questions about notice, consent, proportionality, privacy and the collection of sensitive biometric information.

What people did

  1. Investigated the practice before launching the campaign. CHOICE contacted 25 major Australian retailers and reviewed their privacy policies to establish which companies were using facial recognition.
  2. Visited stores and checked how customers were being notified. The investigation examined signage and whether shoppers were likely to understand that biometric information was being collected.
  3. Published the findings nationally. CHOICE identified the retailers using the technology and explained how facial recognition created biometric faceprints.
  4. Asked supporters to respond. CHOICE surveyed and mobilised its consumer supporter base around the issue.
  5. Filed a formal complaint with the privacy regulator. CHOICE asked the Office of the Australian Information Commissioner to investigate potential breaches of the Privacy Act.
  6. Set out specific legal concerns. The complaint focused on notice and consent, the collection of sensitive information and whether the use of facial recognition was proportionate to the stated purpose.
  7. Maintained public scrutiny while the regulator considered the issue. Consumer pressure and media reporting continued after the initial investigation.

Outcome

The Good Guys paused its facial recognition trial in June 2022. Bunnings and Kmart also paused their systems amid the public controversy.

The OAIC opened formal investigations into Bunnings and Kmart, explicitly stating that the investigations followed CHOICE's reporting.

In September 2025, the Privacy Commissioner found that Kmart's use of facial recognition to address refund fraud breached Australian privacy law. That determination is under review in the Administrative Review Tribunal as of 2026.

Bunnings' case produced a mixed later result. In February 2026, the Tribunal found that Bunnings could rely on a legal exception for its limited use of facial recognition to address serious repeat offending, but upheld findings that Bunnings had breached privacy requirements relating to transparency, notification and privacy governance.

The campaign therefore succeeded in stopping the original retail deployments while they were investigated, triggering national regulatory action and establishing new privacy guidance for facial recognition systems.

Sources

Clearview AI Facial Recognition Database

The OAIC's original determination remains in force.

What was happening

Clearview AI built a facial recognition database by scraping billions of facial images from social media and other publicly accessible websites.

The system included images and biometric templates relating to people in Australia, collected without those individuals directly providing their information to Clearview.

What was done

  1. The Australian privacy regulator opened an investigation. The Office of the Australian Information Commissioner investigated Clearview AI jointly with the United Kingdom's Information Commissioner's Office.
  2. Investigators examined the collection method itself. The case focused on the mass scraping of facial images and conversion of those images into biometric templates.
  3. Applied existing privacy law to the technology. The investigation considered consent, fair collection, notification, accuracy and Clearview's privacy compliance systems.
  4. Made formal findings rather than issuing voluntary guidance. The Australian Information Commissioner determined that Clearview had breached multiple Australian Privacy Principles.
  5. Ordered a specific remedy. The regulator required Clearview to stop collecting facial images and biometric templates from people in Australia and to destroy Australian data already collected.
  6. Defended the determination when Clearview challenged it. Clearview commenced proceedings in the Administrative Appeals Tribunal but later withdrew from the proceedings.

Outcome

The OAIC's original determination remains in force.

Clearview AI is required to cease collecting facial images and biometric templates from individuals in Australia and delete the Australian images it had previously collected.

Sources

7-Eleven Facial Recognition Customer Survey System

7-Eleven stopped collecting facial images and faceprints during the investigation and destroyed the facial images it held.

What was happening

7-Eleven installed customer-feedback tablets with built-in cameras across approximately 700 Australian stores.

When customers completed surveys, the system captured facial images and generated algorithmic biometric representations known as faceprints. These were used to identify repeat survey responses and contribute to demographic analysis.

Around 1.6 million surveys were completed during the first 10 months of the program.

What was done

  1. The privacy regulator investigated the system. The OAIC examined how facial images and biometric faceprints were being collected.
  2. Investigators assessed whether biometric collection was actually necessary for the stated purpose.
  3. The regulator examined whether customers had received adequate notice and provided valid consent.
  4. The case focused on the collection of sensitive biometric information, not only what happened to the data afterwards.
  5. The Privacy Commissioner issued a formal determination that the collection interfered with customers' privacy.
  6. The regulator required destruction of the biometric data.

Outcome

7-Eleven stopped collecting facial images and faceprints during the investigation and destroyed the facial images it held.

The Privacy Commissioner also ordered the company to destroy the remaining faceprints.

Sources

#NotMyDebt and the Robodebt Automated Debt System

In November 2019, the government stopped raising Robodebts based solely on income averaging.

What was happening

The Australian Government's Robodebt program used automated data-matching and income averaging to identify alleged social security debts.

Annual Australian Taxation Office income information could be averaged across fortnightly periods, producing debts that individuals were then expected to disprove.

Robodebt remains one of Australia's most important examples of organised opposition to harmful automated government decision-making.

What people did

  1. People affected began documenting their experiences publicly. The volunteer #NotMyDebt campaign created a website where people could anonymously record what had happened to them.
  2. Built a large evidence base from individual cases. By 2019, #NotMyDebt had published more than 1,200 stories from people affected by the compliance system.
  3. Provided practical information to people challenging debts. The campaign shared resources explaining reviews, complaints and the operation of the system.
  4. Used social media to connect people experiencing the same problem. This turned dispersed individual cases into a visible national issue.
  5. Helped people participate in parliamentary inquiries. #NotMyDebt produced step-by-step guides, templates, posters and printable booklets explaining how to make submissions to Senate inquiries.
  6. Made its own formal submissions. #NotMyDebt submitted evidence to parliamentary inquiries based on the cases it had collected.
  7. Community organisations, academics, legal centres and advocates maintained sustained public scrutiny. This contributed to Ombudsman investigations, Senate inquiries and ongoing media coverage.
  8. Victoria Legal Aid identified suitable test cases. It challenged the legal basis of the system in the Federal Court.
  9. The legal challenge targeted the mechanism at the centre of Robodebt. Lawyers argued that debts could not lawfully be created solely by averaging annual ATO income across fortnightly social security periods.
  10. A test case forced a legal concession. In Deanna Amato's case, the Commonwealth conceded that raising the debt using income averaging was unlawful.
  11. A class action followed. Gordon Legal commenced class-action proceedings on behalf of people affected by Robodebt.
  12. Parliamentary, legal and community scrutiny continued after the system ended. The issue ultimately led to a Royal Commission into the Robodebt Scheme.

Outcome

In November 2019, the government stopped raising Robodebts based solely on income averaging.

The Commonwealth conceded in the Federal Court that the central averaging method used in Deanna Amato's case was unlawful.

Hundreds of thousands of affected debts were later refunded or reduced, class-action proceedings produced compensation, and the Robodebt Royal Commission examined how the scheme was designed, implemented and maintained.

Sources

Fraser Coast Data Centre Water and Power Safeguards, Queensland

On 17 September 2026, Fraser Coast Regional Council voted to require future data-centre developments to provide their own water and power arrangements, with a particular focus on preventing large facilities from relying on local drinking-water supplies.

What was being addressed

With large AI and data-centre developments attracting increasing interest in Queensland, Fraser Coast Regional Council considered how future facilities could affect local electricity and drinking-water supplies.

What was done

  1. A councillor brought a formal motion to council.
  2. The issue was framed around identifiable infrastructure impacts, particularly the protection of drinking-water supplies and pressure on electricity infrastructure.
  3. The motion proposed conditions for future data centres rather than waiting for an individual project dispute.
  4. Councillors debated the resource implications through the formal local-government process.
  5. Council unanimously supported the motion.

Outcome

On 17 September 2026, Fraser Coast Regional Council voted to require future data-centre developments to provide their own water and power arrangements, with a particular focus on preventing large facilities from relying on local drinking-water supplies.

This is an example of a successful pre-emptive policy intervention rather than the cancellation of an individual development.

Source