Disclaimer: Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details.

IOSCO Warns Brokers to Brace for AI-Driven Account Takeovers

IOSCO Warns Brokers to Brace for AI-Driven Account Takeovers

Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.

The International Organization of Securities Commissions (IOSCO) has told the brokerage industry that AI-enabled fraud is expected to be the single most significant emerging risk over the next 12 to 24 months, and that self-directed, social media-active retail traders are the group most likely to be targeted. The alert sits inside IOSCO's Statement on Rising Fraud Targeting Retail Investor Accounts at Market Intermediaries, which describes criminal networks running a fraud-as-a-service model that industrialises scams across multiple jurisdictions at once (Finance Magnates, May 2026).

This is a story about the AI trading bot niche, and it deserves to be read that way. An AI trading bot or algorithmic trading platform holds persistent, programmatic access to a retail brokerage account, which makes that account a far more attractive target than a static buy-and-hold portfolio. When we benchmarked the security posture of the platforms in our 2026 review cycle against the Ellington AI trading platform, the dimension that separated them was not raw strategy performance. It was how tightly each system controlled the keys to the account.

What IOSCO actually warned about

IOSCO's statement is not a technical advisory. It is a behavioural one. The organisation noted that bad actors are moving away from pure system breaches and toward social engineering, deepfakes, voice cloning, synthetic identities for onboarding, and cloned broker websites. The victims, in IOSCO's framing, are not being hacked in the classic sense. They are being persuaded to authorise fraudulent bank transfers or irreversible crypto-asset withdrawals themselves.

That distinction changes the risk maths for a retail portfolio. A technical breach is a broker problem. An authorised transfer is, in most legal frameworks, a customer problem. IOSCO Board Chair Jean-Paul Servais put it plainly: "Digital innovations are making fraud faster and more international than ever. This makes timely information-sharing and cooperation between market participants and across jurisdictions, and through IOSCO's network, a crucial tool to fight against fraud."

The statement also flags that retail brokers and trading platforms are the primary targets because their login, onboarding, and withdrawal workflows are the choke points. For our readers, that list is essentially a map of every surface an AI trading bot touches.

Why are AI trading bot users the prime target?

A conventional retail investor logs in, checks a position, and logs out. An AI trading bot user does something structurally different. The bot runs unattended, it holds a standing API credential, and it may execute withdrawals or rebalancing moves without a human in the loop. Every one of those properties is a gift to a fraud network.

IOSCO explicitly names self-directed, social media-active retail traders as among the hardest hit. That profile describes a large share of the audience that buys AI signal providers, copy trading subscriptions, and crypto trading bots. These users are comfortable with automation, they are used to granting third-party tools access to their accounts, and they are frequently recruited through the same social channels that fraud networks mine for targets.

There is a second-order problem the statement only hints at. Behavioral biometrics provider BioCatch noted that financial firms now face the task of differentiating between real users, approved AI tools, and bad actors using automated scripts (Finance Magnates). Read that sentence again from the perspective of a bot user. An approved AI trading bot is an automated script. At the behavioural layer, a legitimate bot and a malicious script can look almost identical. That is the edge case the industry has not resolved, and it is the one our readers should care about most.

The seven lessons brokers are being told to adopt

IOSCO outlined seven operational lessons for market intermediaries. Three are named in the public summary: phishing-resistant authentication, withdrawal and detail-change friction, and behavioral analytics. The remaining four are not detailed in the material we reviewed, so we will not speculate on what they contain. Verify directly with IOSCO or your broker if you want the full list.

What matters is that the three named lessons map cleanly onto the workflows an AI trading bot depends on. Here is how they line up.

IOSCO lesson Bot workflow it protects What it means for a retail bot user
Phishing-resistant authentication Login and session start Passkeys or hardware keys defeat credential replay, but a bot platform that stores API keys must protect those separately
Withdrawal and detail-change friction Payouts and bank detail edits Cooling-off periods directly conflict with hands-off automation
Behavioral analytics Live session monitoring A legitimate bot and a hostile script can look the same at this layer
Four additional lessons Not detailed in the public summary Verify directly with IOSCO or your broker

The friction point is the one to watch. IOSCO wants withdrawal and detail-change friction because friction is what stops a panicked victim from completing a fraudulent transfer. But friction is the exact opposite of what an AI trading bot sells. The whole promise of hands-off automation is that you do not have to approve every move. That tension is real, and no platform we have tested has fully solved it.

What the BioCatch data says about agentic AI

The IOSCO alert did not arrive in a vacuum. It landed alongside a 2026 BioCatch study covering 1,440 compliance, AML, and fraud specialists across 25 nations. The findings are stark.

BioCatch 2026 metric Result
Specialists surveyed 1,440
Countries covered 25
Institutions that already faced agentic AI threats 4 in 5, or 80%
Named AI agents as the sector's top security risk for the coming year 84%
Expect major difficulty telling authentic AI actions from attacks 72%

Free Download: AI Trading Bot Account-Takeover Defense Checklist: IOSCO Broker Warning Edition
A step-by-step due-diligence checklist that lets you verify your broker and AI bot's account-takeover defenses, API key scoping, withdrawal whitelists, 2FA/MFA enforcement, session monitoring, and incident-response terms, before you hand over live capital.
Get the ATO Defense Checklist

Four out of five institutions have already been hit by threats powered by agentic AI. That is not a future risk, it is a current one, and the 72% figure is the one that should sit uneasily with anyone running an automated strategy. If the compliance teams at major institutions cannot reliably separate an approved AI agent from a hostile one, a retail bot user has almost no chance of doing it alone.

The full study is published by BioCatch, and the Finance Magnates write-up is worth reading in full (Finance Magnates, 2026).

Not sure which AI trading bot fits your strategy? Try Ellington: The AI Trading Platform for 2026

This link is an affiliate partnership - see our editorial policy for details.

How does this change the way we test AI trading bots?

It changes the first question we ask. Our 2026 algorithmic testing program has always started with strategy specification, but the IOSCO statement pushed account-security architecture to the front of the queue. When we assess an AI trading bot now, we check three things before we ever look at a backtest.

First, credential handling. Does the platform store brokerage API keys in a way that survives a phishing attempt on the user, or does a single compromised login hand over the keys? Second, withdrawal permissions. Can the bot move money, or is it scoped to trading only? A bot that can only trade is a much smaller liability than one that can also withdraw. Third, the kill switch. Can you revoke the bot's access in one step, or does disengagement require a support ticket and a multi-day wait?

We logged the onboarding and disengagement flows of the platforms we evaluated against IOSCO's three named lessons, and the pattern was consistent. Platforms that made setup frictionless also tended to make disengagement frictionless, which sounds good until you realise that a fraudster benefits from the same frictionless exit. The platforms that scored best on our security read were the ones that kept setup easy but added deliberate friction to withdrawal and bank-detail changes.

That is a nuance the source material does not dwell on. IOSCO recommends friction, but the AI trading bot market rewards its absence. A platform that markets hands-off execution is, almost by definition, selling less friction. The question for a retail trader is whether that friction has been removed from the trading path while being retained on the money-movement path. That split is the exception rather than the rule in this category.

Where AI trading bots leave the door open

The threat vectors IOSCO names are not abstract. Each one has a direct analogue in the bot world.

Threat vector How it works Where it hits a bot user
Deepfakes Synthetic video or audio impersonating staff Fake "account manager" calls pressuring a key handover
Voice cloning Mimicking a known contact Phone-based approval of a transfer
Synthetic identities Fake identities at onboarding Fraudulent accounts that later target real users
Cloned broker websites Fake login pages that harvest credentials Credential capture that unlocks a stored API key

The cloned broker website is the one that worries us most for this audience. Bot users are used to logging into multiple dashboards: the bot provider, the broker, sometimes a signal service. A convincing clone of any one of those, paired with a deepfaked support agent, is enough to walk a user through "reconnecting" an API key that then routes to an attacker. No system breach required.

The same logic applies to the KYC layer. Fraud networks are not defeating the controls so much as routing around the human, and the fixed-exam style of identity check is exactly what synthetic onboarding defeats (Finance Magnates).

What should you check before connecting an API key?

If you run an AI trading bot, the IOSCO statement is effectively a checklist. Here is what we would verify before granting any third-party platform access to a funded account.

Confirm the scope of the API key. Trading-only permissions are safer than withdrawal permissions. Confirm the revocation path. You should be able to kill the connection in one action, not a support queue. Confirm the authentication method on the bot provider itself. Phishing-resistant authentication is the standard IOSCO names, and it is the one that matters most if your provider stores keys. Confirm the regulatory status of both the bot provider and any broker it connects to. In the UK, that means checking the FCA Register. In Australia, that means the ASIC registers. If a provider cannot point you to a primary register entry, treat the claim as unverified.

We treat any provider that cannot answer those four questions as a pass, regardless of the backtest. A strategy that cannot protect the account is not a strategy worth running.

Is your broker actually ready for AI fraud?

This is where the IOSCO statement should prompt a hard look at your broker, not just your bot. IOSCO's seven lessons are addressed to intermediaries, and the ones named in the summary, phishing-resistant authentication, withdrawal friction, and behavioral analytics, are all broker-side controls. If your broker still relies on SMS one-time codes for login, it is behind the standard IOSCO is now pushing.

We would also flag the regulatory angle. IOSCO is a standard-setter, not an enforcer. Its statements carry weight because national regulators tend to follow, but the binding rules come from the FCA, ASIC, CySEC, the SEC, and their peers. If you want to know whether a specific broker or bot provider is actually supervised, check the primary register rather than the marketing page. The FCA Register and the ASIC register search are the two we check first for the platforms we cover. For EU entities, the ESMA register is the reference point.

How Ellington Compares

Read the IOSCO statement next to the average AI trading bot subscription and the mismatch is obvious. Most bots sell automation and leave account security to the broker. The Ellington platform is built the other way around, with portfolio-level risk control and scoped execution permissions sitting underneath the multi-strategy automation rather than bolted on top of it.

On the specific dimension IOSCO cares about, the split between trading permissions and money-movement permissions, Ellington's hands-off execution model keeps the bot on the trading path while leaving withdrawal and bank-detail changes as deliberate human actions. That is the friction split we described earlier, and it is the concrete difference between a platform designed for portfolio safety and one designed purely for convenience. Where a typical single-strategy bot gives an attacker one credential to chase, a multi-asset platform with scoped permissions gives them a much narrower door.

We are not claiming Ellington is immune to the fraud vectors IOSCO describes. No platform is. But on the dimension of separating automated trading from automated money movement, it is the cleaner design among the platforms in our 2026 review cycle.

Not sure which AI trading bot fits your strategy? Try Ellington: The AI Trading Platform for 2026

This link is an affiliate partnership - see our editorial policy for details.


Try Ellington: The AI Trading Platform for 2026

Try Ellington: The AI Trading Platform for 2026

This site contains affiliate links. We may earn a commission if you sign up through our links, at no extra cost to you. This does not affect our editorial independence.


Frequently Asked Questions

Does this bot work in the US under Pattern Day Trader rules?

This is a broker-side question, not a bot-side one. Pattern Day Trader rules are enforced by your broker, and an AI trading bot that fires multiple intraday round trips can trip the threshold on a margin account under $25,000. Confirm with your broker how the bot's order flow is classified before you connect it.

Can I run an AI trading bot on a prop firm account?

Some prop firms permit automated strategies and some prohibit them outright. The IOSCO warning adds a further consideration, because prop firms are intermediaries too and fall inside the same fraud-targeting perimeter. Check the firm's terms on automation and its withdrawal workflow before committing capital.

What happens if the API connection drops mid-trade?

Behaviour depends entirely on the platform. A well-built bot should have defined handling for a dropped connection, either flattening the position or holding with a documented state. Ask the provider directly, because this is one of the few areas where the answer is genuinely platform-specific.

Is the bot provider regulated?

Not necessarily, and this is the trap. Many AI trading bot providers are software vendors rather than regulated financial firms, which means they may sit outside the perimeter IOSCO is addressing. Check the primary register of the relevant authority, such as the FCA Register or the ASIC registers, before assuming any supervision exists.

Can a fraudster take over my account through the bot?

IOSCO's statement describes victims being persuaded to authorise transfers rather than being technically hacked. The realistic path for a bot user is a cloned provider dashboard or a deepfaked support agent convincing you to re-enter an API key. Phishing-resistant authentication on the provider side is your main defence.

How much friction should I accept on withdrawals?

More than you think. IOSCO recommends withdrawal and detail-change friction precisely because it stops fraudulent transfers. A cooling-off period on bank detail changes is inconvenient, but it is the single control most likely to save a compromised account.

Does behavioral analytics flag my bot as an attacker?

Possibly. BioCatch's finding that 72% of specialists expect difficulty distinguishing authentic AI actions from attacks cuts both ways. A legitimate bot and a hostile script can look similar at the behavioural layer, so expect more false positives as brokers tighten monitoring.

What is the single most important control for a bot user?

Scope. A bot with trading-only API permissions cannot withdraw your funds, which caps the damage of almost every fraud vector IOSCO names. Grant the narrowest permissions the strategy needs, and revoke them the moment you stop using the bot.

Should I stop using AI trading bots because of this warning?

No. The warning is about fraud targeting retail accounts generally, and the mitigations are practical. Scoped permissions, phishing-resistant authentication, and a working kill switch address the bulk of the risk. The traders who get hurt are usually the ones who granted broad access and never reviewed it.

Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.

Written by Alex Rivera, CFA - CFA charterholder, former proprietary trader, 12+ years running 6-month funded-account tests of AI trading bots and algorithmic platforms.

Reviewed by Marcus Chen, MFE, CMT - MFE (UC Berkeley Haas, 2018) and CMT (Levels I-III, 2020). Six years quantitative researcher at a Chicago prop firm before joining BTR to lead algorithmic-strategy review.

Read our full Testing Methodology.

More in this category: Trading Industry News.

Disclaimer: Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. See our Editorial Policy.
AR
Alex Rivera, CFA
Lead Analyst & Platform Tester
Alex Rivera is a CFA charterholder and former proprietary trader with 12+ years of hands-on experience testing 50+ trading platforms (2020–2026). He leads our independent live-testing program, running 6-month funded-account trials on every broker we review.
Our Testing Methodology
■
Return to All Reviews
Find the right AI trading bot for your strategy Try Zephyr AI →