AI trading regulation in Australia: what ASIC requires
AI trading is legal in Australia and regulated through existing licensing and market integrity rules, not through an AI statute. An AFS licensee must provide services efficiently, honestly and fairly whether or not a model is involved. Market participants operate under the Market Integrity Rules, which ASIC proposed to modernise for algorithmic and AI-driven trading in CP 386 on 27 August 2025 — a consultation, not law. Retail CFD leverage is capped by a product intervention order running to 23 May 2027, and 68% of retail CFD investors still lost money in FY2024.
Published 27 August 2026 · Updated 28 August 2026 · AI Trading Book Editorial · Reading time about 15 minutes
- Australia has the most automated market of the three we cover. ASIC estimates algorithmic trading at roughly 85% of listed equities and 94% of SPI 200 futures.
- CP 386 is the most significant proposed change since 2012 — a Trading Algorithm definition, immediate-suspension controls, and one rulebook across securities and futures. Still a proposal.
- REP 798 found adoption outrunning governance across 624 use cases at 23 licensees.
- The CFD intervention worked and did not solve the problem. Net losses fell 91%; 68% of retail CFD investors still lost money in FY2024.
- Wholesale classification removes those protections. It is a reduction in protection, not a status upgrade.
Is AI trading legal in Australia?
Yes. No Australian statute bans AI trading, and none creates a separate legal category for it. The determining question is the same as in the United Kingdom: what activity is being performed, and for whom. Running an automated strategy on your own account through an Australian-licensed broker is not the provision of a financial service and requires no licence.
Obligations attach at three points. A person or business providing financial services to others generally needs an Australian Financial Services Licence. A firm that is a participant of a licensed market — ASX, Cboe Australia, NSX or SSX — operates under the ASIC Market Integrity Rules, which govern how orders reach the market. And anyone dealing with retail clients in leveraged derivatives is subject to the product intervention orders that cap leverage and ban binary options.
The practical mistake is assuming that "AI" moves a product outside this framework or into a special category. It does neither. A robo-adviser is an adviser; an automated order router is a trading system; a signal service that names specific products to specific people is providing financial product advice.
The AFSL regime and section 912A
The Australian Financial Services Licence is the gateway obligation. A business that deals in financial products, provides financial product advice, makes a market or operates a registered scheme generally needs one, and the licence specifies which authorisations it holds. A licence to deal does not authorise advice; a licence covering derivatives does not automatically cover securities.
Section 912A of the Corporations Act 2001 carries the general obligations, of which the best known is the requirement to do all things necessary to ensure that financial services are provided efficiently, honestly and fairly. Australian courts have treated that phrase as a compound standard rather than three separate tests, and it is deliberately technology-neutral. A licensee that delivers advice through a model is subject to it exactly as one that delivers advice through a person.
Two further s912A obligations matter for AI specifically: the requirement to have adequate resources and risk management systems, and the requirement to ensure representatives are adequately trained and competent. A firm deploying a model it cannot explain has a defensible problem on both counts, and that is close to what ASIC found when it looked.
REP 798: adoption outrunning governance
ASIC published REP 798 on 29 October 2024, reporting a review of AI use across 23 licensees covering 624 individual use cases. It is the most substantial Australian evidence base on how financial firms actually deploy AI, and its central finding is a governance gap rather than a technology failure: licensees were adopting AI faster than they were updating their risk and compliance frameworks.
Two supporting figures give the shape of the adoption curve. Of the licensees reviewed, 61% intended to increase their AI use. And 92% of the generative AI use cases identified dated from 2022 to 2023 — meaning that at the time of review, almost the entire generative AI footprint in the sample was under two years old, deployed into governance frameworks written before it existed.
This mirrors what the FCA found in its 2025 review of algorithmic trading controls, discussed on our UK regulation page: the controls tend to be present, and the understanding tends to lag. Both regulators arrived at supervisory expectations rather than new rules.
The Market Integrity Rules
Market participants operate under two rulebooks: the ASIC Market Integrity Rules (Securities Markets) 2017 and the ASIC Market Integrity Rules (Futures Markets) 2017. These govern how a participant's systems connect to a market, what controls sit in front of an order, and what a participant must do when something goes wrong. RG 241 on electronic trading provides supporting guidance.
The current securities rules distinguish between designated trading representatives and automated order processing — a split that dates from an era when a human keying an order and a machine generating one were meaningfully different activities. The futures rules contain neither concept. That divergence is the main thing CP 386 sets out to fix.
CP 386: what ASIC has proposed
CP 386, published 27 August 2025, proposes amendments to both Market Integrity Rules and is the most significant change to the electronic trading rules since 2012. ASIC's stated objectives are to modernise the trading systems obligations to reflect current practice, technology and risks; to drive consistency between the securities and futures rulebooks; and to align the framework with IOSCO principles and international practice on algorithmic trading. Submissions were lodged through October 2025. It is a consultation paper: nothing in it is currently binding.
The proposed Trading Algorithm definition
CP 386 would insert a new defined term, Trading Algorithm, into both rulebooks — described in the draft as a computer algorithm that automatically determines, with limited or no human intervention, one or more parameters of an order. The definition therefore captures systems that decide the substance of an order rather than merely the mechanics of submitting and managing it, and it captures a machine-learning model exactly as it captures a rules-based one.
Alongside it, the proposals would replace the designated trading representative and automated order processing distinction with a single concept of Trading Systems, defined broadly as any system for submitting trading messages into a trading platform. The effect is that the obligations would apply regardless of how the order was generated — the same technology-neutral move the UK made when it onshored RTS 6.
Immediate suspension controls
Among the new rules proposed for Trading Algorithms is a requirement for participants to have controls enabling immediate suspension of algorithmic activity. This is the Australian equivalent of the kill functionality that article 12 of UK-onshored RTS 6 already requires. If adopted, it would close one of the clearer gaps between the two regimes.
Industry submissions raised proportionality concerns, particularly around a proposal for real-time monitoring across the board, on the grounds that the systems and staffing cost would be disproportionate to the risk for many participants. Submissions also sought a long transition period. What ASIC does with that feedback is not yet known.
| Area | Australia — proposed in CP 386 | United Kingdom — in force |
|---|---|---|
| Definition of the regulated object | New "Trading Algorithm" term; broad "Trading Systems" concept | Algorithmic trading defined in MAR 7A |
| Emergency stop | Controls enabling immediate suspension | Kill functionality, RTS 6 article 12 |
| Testing before deployment | Development, testing and monitoring controls proposed | Conformance testing, RTS 6 article 6 |
| Periodic review | Governance arrangements proposed | Annual self-assessment and validation, RTS 6 article 9 |
| Adverse-condition testing | Not separately identified in the proposals reviewed | Stress testing, RTS 6 article 10 |
| Status | Consultation; submissions lodged October 2025 | Binding rules |
How automated is the Australian market?
More than most. Publishing CP 386 on 27 August 2025, ASIC estimated that algorithmic trading comprises approximately 85% of all trading in Australian listed equities markets, about 94% of SPI 200 futures trading, and about 46% of three-year bond futures. Those are the regulator's own estimates, published to justify the rule change.
Two cautions on reading these numbers. They measure algorithmic trading in the broad sense — any automated order generation — not AI or machine learning specifically, and the great majority of that volume is execution algorithms rather than predictive models. And they are participant-side estimates of market activity; they say nothing about how much of it is profitable, or about retail participation.
The retail CFD product intervention order
Most retail AI trading in Australia happens on leveraged products, which makes the product intervention order the rule that most affects retail outcomes. ASIC's order took effect on 29 March 2021, announced in media release 20-254MR, and was extended by 22-082MR to run until 23 May 2027. A separate order banning the issue and distribution of binary options to retail clients took effect on 3 May 2021.
| Requirement | Limit |
|---|---|
| Leverage — major currency pairs | 30:1 |
| Leverage — minor currency pairs, gold, major equity indices | 20:1 |
| Leverage — other commodities and minor equity indices | 10:1 |
| Leverage — crypto-asset CFDs | 2:1 |
| Margin close-out | At 50% of total initial margin |
| Negative balance protection | Required; retail client cannot lose more than the account balance |
| Binary options | Prohibited for retail clients from 3 May 2021 |
| Order in force | 29 March 2021 to 23 May 2027 |
The 2:1 crypto cap is the one most often missed by people building automated crypto strategies. It applies to crypto-asset CFDs offered to Australian retail clients by an AFS licensee — not to spot crypto on an offshore exchange, which is a different product with different protections and, in most cases, none of these.
Did the intervention work?
On ASIC's own measurement, substantially. REP 724 reported that following the order, aggregate retail client net losses fell by 91% and the number of loss-making retail accounts fell by 51% per quarter. That is one of the larger documented effects of a retail conduct intervention anywhere.
It did not make retail CFD trading profitable. REP 828, published 20 January 2026, found that 68% of retail CFD investors lost money in FY2024. Both facts are true at once: the intervention removed a large amount of harm, and the residual harm is still the majority outcome.
Retail versus wholesale clients
The single most consequential classification in Australian financial services is whether you are a retail or a wholesale client, because wholesale status removes most of the protections described above — including the product intervention order's leverage caps, much of the disclosure regime, and access to some dispute-resolution avenues.
The tests include a product value threshold, an accountant's certificate confirming net assets or gross income above set levels, and professional and sophisticated investor categories. Firms marketing high-leverage automated strategies sometimes present wholesale certification as a way to unlock better terms. It is more accurately described as opting out of consumer protection: the leverage available goes up, and the regulatory floor beneath you disappears.
The regulatory timeline
Five dated actions define the current Australian position. Reading them in order shows a regulator moving from consumer protection in leveraged products towards market-integrity rules for automation.
| Date | Action | Effect |
|---|---|---|
| 29 March 2021 | Product intervention order takes effect (20-254MR) | Retail CFD leverage caps, margin close-out, negative balance protection |
| 3 May 2021 | Binary options ban takes effect | Issue and distribution to retail clients prohibited |
| — | REP 724 | Measures the intervention: net losses down 91%, loss-making accounts down 51% |
| 2022 (22-082MR) | Product intervention order extended | Runs to 23 May 2027 |
| 29 October 2024 | REP 798 | 624 AI use cases at 23 licensees; adoption outpacing governance |
| 27 August 2025 | CP 386 published | Proposes Trading Algorithm definition and immediate-suspension controls |
| October 2025 | Submissions lodged on CP 386 | Industry raises proportionality and transition concerns |
| 20 January 2026 | REP 828 | 68% of retail CFD investors lost money in FY2024 |
AI-branded fraud is the larger retail risk
Measured by documented consumer harm, fraud outweighs everything else on this page. ASIC coordinated the removal of 11,964 phishing and investment scam websites during 2025, against 6,270 in the preceding twelve-month period — a 90% increase — along with more than 1,100 social media investment scam advertisements. More than 25,000 sites have been removed since 2023.
"AI" is the dominant marketing term in that material. ASIC has documented fabricated brands including "Quantum AI" and "Deep Blue", the latter advertising returns of 85% and 374%. Our scams page covers the patterns and the register checks in detail.
What to verify before using a platform
Four checks resolve most of the risk, and all are free.
- Look up the AFS licence on ASIC Connect's professional registers. Confirm the licence exists, is current, and that its authorisations cover what is being offered — dealing, advice, or both.
- Check the entity name against the one on the register, not the trading or brand name. Impersonation of licensed entities is a documented pattern.
- Confirm you are being treated as a retail client unless you have deliberately chosen otherwise and understand what you gave up.
- Check the external dispute resolution membership. An AFS licensee dealing with retail clients must be a member of the Australian Financial Complaints Authority. No AFCA membership means no independent complaint route.
What we could not establish
Three things a reader might expect are absent, because no primary source was found.
- Any ASIC assessment of whether AI trading tools are profitable. ASIC's published work addresses governance, market integrity and consumer outcomes. It does not evaluate tool performance, and figures circulating in secondary sources claiming an ASIC finding on loss rates for algorithmic bots could not be traced to any ASIC publication. Data not found.
- A count of Australian retail traders using AI tools. No official series exists. Data not found.
- A breakdown of the 85% algorithmic share by participant type. ASIC published the aggregate estimate without splitting institutional from retail-originated flow. Data not found.
Key takeaways
- Nothing AI-specific is currently in force. The obligations are the AFSL regime, s912A, the Market Integrity Rules and the product intervention orders. CP 386 would change the third of those; it has not yet.
- Australia is the most automated of the three markets we cover, at roughly 85% of listed equities and 94% of SPI 200 futures on ASIC's estimate — which is why the rule modernisation started here first.
- The proposed Trading Algorithm definition is technology-neutral by construction. It turns on whether a system determines order parameters with limited human intervention, which captures machine learning without naming it.
- The governance gap is the documented problem. REP 798 found adoption running ahead of risk frameworks across 624 use cases, the same pattern the FCA found in the UK.
- Leverage caps do more for retail outcomes than any AI rule. Net losses fell 91% after the intervention, and 68% of retail CFD investors still lost money in FY2024.
- Wholesale certification is a downgrade in protection. Treat an offer to certify you as wholesale as a warning, not a benefit.
Frequently asked questions
Is AI trading legal in Australia?
Yes. No Australian law bans AI trading and there is no AI-specific trading statute. Trading your own money through a licensed broker using an AI tool is not a regulated activity. Regulation attaches to financial services provided to others, and to market participants trading on licensed markets, regardless of whether a model is involved.
Does ASIC have AI-specific trading rules?
Not in force. ASIC regulates AI through existing obligations: the AFSL licensing regime, the section 912A obligation to provide services efficiently, honestly and fairly, and the Market Integrity Rules. CP 386, published 27 August 2025, proposes amendments addressing algorithmic and AI-driven trading directly, but it is a consultation, not law.
What is ASIC CP 386?
CP 386 is a consultation paper published on 27 August 2025 proposing amendments to the ASIC Market Integrity Rules for Securities Markets and Futures Markets. It would modernise the trading systems obligations, introduce a definition of Trading Algorithm, require controls enabling immediate suspension of algorithmic activity, and harmonise the two rulebooks.
How much Australian trading is algorithmic?
ASIC estimates approximately 85% of all trading in Australian listed equities markets, about 94% in SPI 200 futures and about 46% in three-year bond futures. These figures were published with CP 386 on 27 August 2025 and cover automated order generation generally, not AI specifically.
What would the CP 386 kill switch requirement mean?
CP 386 proposes that participants must have controls enabling immediate suspension of a Trading Algorithm. It is a proposal, not a rule in force. The equivalent obligation already exists in the United Kingdom under article 12 of onshored RTS 6, which requires the ability to cancel unexecuted orders immediately.
What did ASIC REP 798 find about AI use?
REP 798, published 29 October 2024, reviewed 624 AI use cases across 23 licensees. It found 61% planned to increase AI use, and 92% of generative AI use cases dated from 2022 to 2023. Its central warning was that licensees were adopting AI faster than they were updating risk and compliance frameworks.
What leverage limits apply to Australian retail CFD traders?
ASIC's product intervention order caps retail CFD leverage at 30:1 for major currency pairs, 20:1 for minor pairs, gold and major equity indices, 10:1 for other commodities, and 2:1 for crypto-asset CFDs. It also requires 50% margin close-out and negative balance protection. The order took effect 29 March 2021 and runs to 23 May 2027.
Did ASIC's CFD intervention work?
On ASIC's own measurement in REP 724, yes. Aggregate retail client net losses fell by 91% and loss-making retail accounts fell by 51% per quarter. Losses did not disappear: REP 828 found 68% of retail CFD investors still lost money in FY2024.
Are binary options banned in Australia?
Yes, for retail clients. ASIC's product intervention order banning the issue and distribution of binary options to retail clients took effect on 3 May 2021. It is separate from the CFD order, which restricts rather than prohibits.
Do I need an AFSL to run a trading bot?
Not for your own account. An AFS Licence is required to provide financial services to others — dealing, advising, or making a market. Running an automated strategy on your own money through a licensed broker is not a financial service. Managing other people's money or selling personal advice generally is.
What is the difference between retail and wholesale clients?
Wholesale client status removes most retail protections, including the product intervention order's leverage caps and much of the disclosure regime. The tests include a product value threshold, an accountant's certificate of net assets or income, and professional investor categories. It is a significant reduction in protection, not a status upgrade.
Does the obligation to act efficiently, honestly and fairly cover AI?
Yes. Section 912A of the Corporations Act 2001 requires an AFS licensee to provide financial services efficiently, honestly and fairly. The obligation is technology-neutral and applies to services delivered through an AI system exactly as it applies to services delivered by a person.
Are AI trading bot vendors licensed in Australia?
Often not, and availability proves nothing. Selling general-purpose software is usually outside the licensing perimeter, while advising on specific products is inside it. Check ASIC Connect's professional registers and confirm the authorisations cover what is offered. ASIC coordinated the removal of 11,964 scam websites in 2025 alone.
Can I use an offshore AI trading platform from Australia?
Access is usually technically possible, but the protections described on this page do not travel with you. An offshore provider without an AFS licence is not bound by the leverage caps, is not required to hold AFCA membership, and offers no Australian dispute route. Availability in Australia is not authorisation in Australia.
Compiled by AI Trading Book Editorial from ASIC consultation papers, reports and media releases, the Corporations Act 2001 and the ASIC Market Integrity Rules 2017. Every figure carries a named source and a date below. Where a figure could not be traced to a primary source it is marked "data not found" rather than estimated. CP 386 is identified throughout as a consultation, not law. Published 27 August 2026; last updated 28 August 2026. Corrections are logged on the corrections page.
Sources
- ASIC — CP 386, "Proposed amendments to the ASIC market integrity rules: Trading systems and automated trading", 27 August 2025, with draft rules attachments.
- ASIC — "ASIC moves to modernise trading system rules to keep pace with technology and AI", 27 August 2025 — algorithmic share estimates of approximately 85% of listed equities, about 94% of SPI 200 futures and about 46% of three-year bond futures; IOSCO alignment objective.
- ASIC Market Integrity Rules (Securities Markets) 2017 and ASIC Market Integrity Rules (Futures Markets) 2017; RG 241 on electronic trading.
- ASIC — REP 798, 29 October 2024 — 624 AI use cases across 23 licensees; 61% intending to increase use; 92% of generative AI use cases from 2022–23; adoption outpacing risk and compliance frameworks.
- ASIC — REP 828, 20 January 2026 — 68% of retail CFD investors lost money in FY2024.
- ASIC — REP 724 — aggregate retail client net losses down 91%; loss-making retail accounts down 51% per quarter following the product intervention order.
- ASIC — 20-254MR, product intervention order in effect from 29 March 2021; leverage caps 30:1, 20:1, 10:1 and 2:1; 50% margin close-out; negative balance protection.
- ASIC — 22-082MR, extension of the product intervention order to 23 May 2027.
- ASIC — binary options product intervention order, in effect from 3 May 2021.
- ASIC — 26-063MR, 8 April 2026 — 11,964 phishing and investment scam websites removed in 2025 against 6,270 in the preceding twelve months; more than 1,100 social media advertisements; more than 25,000 sites since 2023.
- ASIC — 24-180MR — fabricated "Quantum AI" brand.
- Corporations Act 2001, section 912A — general obligations of AFS licensees, including provision of financial services efficiently, honestly and fairly.
- Law Council of Australia and AFMA — submissions on CP 386, lodged October 2025 — proportionality and transition concerns, cited for industry response only.
- FCA — multi-firm review of algorithmic trading controls, 21 August 2025, and onshored RTS 6 — cited for cross-jurisdictional comparison only.
Informational research only. Nothing on this page is personal financial, legal, tax or investment advice, or a recommendation to trade any instrument. CP 386 is a consultation paper and its proposals are not binding. Regulatory rules change; verify the current position with ASIC before relying on it, and take qualified legal advice where your licensing position is uncertain.