Australian AI Governance Briefing: Week Ending 23 August 2026

Two developments this week, both from the Commonwealth. On Monday 17 August ASIC published a report on AI-generated investment scams, disclosing that it removed more than 19,400 online scams in FY26 — an increase of 182% on the previous year — and warning that a quick online search is no longer sufficient to verify an investment opportunity. The regulator’s point is structural rather than cautionary: generative AI now lets scammers build mutually reinforcing networks of fake platforms, fabricated news articles, synthetic reviews and deepfake endorsements, so the due diligence channel itself returns the scam.

On Tuesday 18 August, Assistant Minister for Science, Technology and the Digital Economy Andrew Charlton delivered the ANU Crawford School of Public Policy Annual Lecture, arguing that Australia should convert its data centre investment boom into domestic AI capability rather than settling for the infrastructure layer. The operationally significant line was a policy signal: Charlton said he and Industry Minister Tim Ayres are working on how to make Expectation 5 of the National Data Centre Expectations — which currently asks large-scale compute providers to offer Australian start-ups, small businesses, researchers and not-for-profits access to compute on favourable terms — a requirement of hyperscalers, in support of the Prime Minister’s proposed National AI Standards.

Parliament sat from Monday to Thursday without AI-specific business.

The week in review

ASIC on AI-generated scam networks

ASIC published media release 26-195MR on 17 August, warning that scammers are using generative AI to create what the regulator described as vast webs of deception designed to defeat the ordinary checks a consumer would run before investing. The mechanism ASIC describes is a network rather than a single fake site: a deepfake video endorsement of a well-known figure leads to a spoofed news article, which links to a fraudulent trading platform carrying fabricated reviews, with each element lending apparent legitimacy to the others. The practical consequence is that a search intended to verify an opportunity surfaces the scam material rather than filtering it out.

The takedown figures are the substance of the release. ASIC removed more than 19,400 online scams in FY26, up 182% on the prior year, covering fake websites, social media advertisements, phishing scams and cryptocurrency investment scams. Removals of cryptocurrency investment scams rose by almost 30%, with 3,106 scams taken down over the year. Since ASIC launched its takedown capability three years ago it has removed more than 33,400 scam websites, social media advertisements and phishing scams.

The regulator identified the most impersonated public figures in FY26 as Anthony Albanese, Jacqui Lambie, Angus Taylor, Tom Piotrowski and Alan Kohler, attributing the list to the National Anti-Scam Centre. ASIC also noted that criminals are increasingly building scams around topical issues in the news, which shortens the interval between a live news cycle and the appearance of scam content exploiting it.

ASIC’s consumer guidance directs people to verify a licence holder’s name and number on ASIC’s professional registers, to check that those details match the business actually promoting the opportunity, and to consult the Moneysmart Investor Alert List. The corresponding message for licensees is the one ASIC has been making since 26-122MR: industry participants are urged to add their AFS licensee website addresses to the Professional Registers Search, so that consumers can distinguish a genuine bank, investment platform or superannuation fund from a clone. For AFS licensees whose brands and licence numbers are being misused, register hygiene is now a consumer-protection control rather than an administrative task.

Speaking to the ABC, ASIC Chair Sarah Court said social media companies have a great deal to answer for in relation to scams, and pointed to the Scams Prevention Framework as the mechanism that will bring platforms into the response. That framing situates the report within the broader shift from regulator-led takedown to distributed obligations across the ecosystem.

Primary sources: ASIC — 26-195MR ASIC warns scammers are using AI to spin vast webs of deception | ABC News — AI deepfake scams as ASIC reports rise in false investment endorsements

Charlton signals a mandatory compute-access obligation on hyperscalers

Assistant Minister Andrew Charlton delivered the ANU Crawford School of Public Policy Annual Lecture on 18 August, under the title Why Australia needs to turn compute into Sovereign AI. The economic argument was that only a small share of AI spending accrues to electricity and the physical data centre, with most of the value flowing further up the stack to chips and servers, models and software, and research and intellectual property. Charlton put the position that if Australia supplies the land, power and buildings while foreign firms own the chips, the models and the customer relationship, the country captures some of the wealth while the largest rents accrue elsewhere. He framed the goal as competing across the great majority of the AI stack rather than being confined to its infrastructure layer.

The regulatory content sat in the second of three propositions. The government’s Expectations of data centres and AI infrastructure developers, released on 23 March 2026, already ask hyperscalers to make compute available to Australian start-ups and to partner with the domestic innovation ecosystem. Expectation 5 specifically asks providers of large-scale compute to enable access for Australian start-ups, innovative small businesses, researchers and not-for-profits on favourable terms. Charlton said that Minister Ayres and he are working on how that expectation can be made a requirement of hyperscalers, to support the Prime Minister’s National AI Standards, and described building the ecosystem around the infrastructure deliberately as making it a condition of the welcome.

That is a change of legal character, not of emphasis. The expectations as published in March are not legally binding, and industry responses at the time proceeded on that basis — Data Centres Australia noted that many of its members were already meeting the expectations voluntarily and without regulatory obligation, while objecting that on-premises data centres were excluded from scope. Converting Expectation 5 into a requirement would fold a compute-access and local-partnership condition into the legislated Australian Standards for AI that the Prime Minister announced on 15 July 2026, alongside the establishment of an Office of AI within the Department of the Prime Minister and Cabinet, with legislation expected in Parliament in early 2027.

For compliance and government affairs teams at hyperscalers, data centre developers and their Australian counterparties, the practical implication is that commercial commitments currently being offered voluntarily — compute access on favourable terms, local engineering and research deployment, supply chain investment, university partnerships and talent attraction — are candidates to become conditions carrying legal consequence. The remaining variables are which entities fall within scope, what “favourable terms” means when it has to be drafted rather than described, and what compliance and enforcement look like. None of that is settled: the speech is a signal of work in progress, not an instrument.

Primary sources: Ministers for DISR — ANU Crawford School of Public Policy Annual Lecture | Crawford School — Hon. Dr. Andrew Charlton, MP: ‘Australia’s AI growth opportunity’ | Ministers for DISR — Press conference announcing Data Centre Expectations

Stories

ASIC reports 182% rise in scam takedowns as AI builds self-verifying scam networks

ASIC published media release 26-195MR on 17 August 2026, reporting that it removed more than 19,400 online scams in FY26, up 182% on the prior year, including 3,106 cryptocurrency investment scams — an increase of almost 30%. The regulator warned that generative AI now enables scammers to build interconnected networks of deepfake endorsements, spoofed news articles, fake platforms and fabricated reviews, so that a consumer’s online search returns the scam material rather than filtering it out. The most impersonated public figures in FY26 included Anthony Albanese, Jacqui Lambie, Angus Taylor, Tom Piotrowski and Alan Kohler, according to the National Anti-Scam Centre. ASIC urged industry participants to add their AFS licensee website addresses to the Professional Registers Search so consumers can verify whether a bank, investment platform or super fund is legitimate.

Source: asic.gov.au

Charlton says government is working to make hyperscaler compute-access expectation a requirement

Assistant Minister for Science, Technology and the Digital Economy Andrew Charlton used the ANU Crawford School of Public Policy Annual Lecture on 18 August 2026 to argue that Australia should convert its data centre investment boom into domestic AI capability rather than remaining at the infrastructure layer. He said that he and Industry Minister Tim Ayres are working on how to make Expectation 5 of the National Data Centre Expectations — which asks providers of large-scale compute to enable access for Australian start-ups, innovative small businesses, researchers and not-for-profits on favourable terms — a requirement of hyperscalers, in support of the Prime Minister’s National AI Standards. The expectations, released on 23 March 2026, are not legally binding. Legislation for the Australian Standards for AI is expected in Parliament in early 2027.

Source: minister.industry.gov.au

---

This briefing was researched and written with AI assistance.