Belgium Blacklists 51 More Fake Trading Platforms, Nearly Matching All of 2025

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.

Belgium's Financial Services and Markets Authority (FSMA) named 51 more fraudulent trading platforms and the websites steering investors to them on Thursday, bringing its 2026 count to roughly 150 — just shy of the 157 platforms it flagged in all of 2025, with three months still to run. For anyone running an AI trading bot, an algorithmic trading platform, or a copy trading service on a funded account, the headline is not really about Belgium. It is about the fact that the fastest-growing category of consumer-facing "automation" in Europe right now is not a strategy at all — it is a wrapper designed to look like one. When we benchmarked a representative basket of these flagged-style domains against the Ellington AI trading platform in our 2026 review cycle, the difference was not performance. It was whether the thing had a verifiable regulatory footprint at all.

What did Belgium's regulator actually publish?

The FSMA's Thursday list is the largest of the four it has published in 2026, following 24 names in March, 47 in April, and 25 at the end of June (FinanceMagnates, May 2026). Ten of the 51 entries are not trading platforms themselves — they are affiliate-style websites that route consumers to fraudulent platforms. Several carry Dutch- or French-sounding names, including Gent Beursveks, Koersaven, Haute Mondrève and Belvue Capitange.

Three platforms — Bonmot Invest, Canford Castle and Fid Europeed — were marked as clones of other firms. Others borrow brand equity from recognizable names: ICBC Markets, SGoldmanIFA, and MT5 Novatrade. That last one is worth pausing on. MetaTrader 5 is a legitimate, widely used execution terminal. A domain that appends "MT5" to a made-up brand is not a MetaTrader broker, and it is not an expert advisor (EA) provider either — it is impersonation of the terminal layer.

The pattern we keep seeing in our own bot due-diligence work is that fraudsters increasingly hide behind the vocabulary of automation. "AI" in a URL, "algo" in a landing page, a fake MT5 login screen — these are costume, not infrastructure. The FSMA noticed the same thing: some of the flagged domains add "ai" to their web addresses, as in ai-gentbeursveks.com and koersaven-ai.com.

The numbers behind the crackdown

Reported losses on fraudulent trading platforms have fallen for three consecutive half-years, from EUR 12.5 million in the second half of 2024 to EUR 8.5 million in the first half of 2026, against EUR 11.9 million a year earlier. That is a genuine decline, and it is worth taking seriously rather than dismissing as a reporting artifact.

The divergence matters. Warning volume is up roughly 60% year over year on entities (98 to 157) while consumer complaints are flat (1,289 to 1,277). Our read: enforcement is getting faster and more surgical, not that the underlying fraud is shrinking. The FSMA fed 245 fraudulent websites into the Belgian Anti-Phishing Shield in 2025, and that Shield redirected visits from 22,973 unique IP addresses away from them. That is supply-side intervention — blocking the funnel before the victim ever funds an account.

Why AI-branded scams are the hard case

A traditional boiler-room scam requires a human on the phone. An AI-branded platform scam requires a landing page, a payment processor, and a plausible-looking dashboard. That is a much lower marginal cost per victim, and it scales.

The FSMA's August warning about fake news sites dressed up as Le Soir, La Libre, HLN and Euronews — funnelling readers to a platform called BitKeltTrade — is the template. In France, brand impersonation accounted for 85.1% of warnings issued since 2010. The impersonation is not of a broker. It is of a news brand, because trust in the media is cheaper to borrow than trust in a regulator.

Here is the editorial point that most coverage of this story misses. The regulatory perimeter is drawn around who holds client money, not around who generates the trade signal. A fraudulent "AI signal provider" that never touches client funds and never executes a trade sits in a gap: it is not a broker, so CySEC or the FCA may not have jurisdiction; it is not a fund, so there is no AIFM registration; and it is not an investment adviser if it disclaims advice. The FSMA can blacklist the domain, but it cannot easily prosecute the entity if the entity does not exist on paper. Retail traders evaluating any AI signal provider should treat "we are not a broker" as a red flag rather than a reassurance.

How do we separate a real bot from a fake one?

Our 2026 algorithmic testing program uses a fixed checklist. It is not glamorous, but it filters out the overwhelming majority of the problem.

First, the provider must be named on a primary regulator register, and we verify it there ourselves rather than trusting a badge on the website. For UK entities that means the FCA Register; for Australian entities, the ASIC AFSL search; for EU entities, the CySEC list or the ESMA register. If a provider claims a licence we cannot find on the primary register, we stop the review. No exceptions.

Second, the strategy must be describable in plain English without reference to proprietary "AI." If the provider cannot tell you what it trades, on what timeframe, with what position-sizing rule, and under what stop condition, there is no strategy to evaluate — only a return chart.

Third, the fee model must be stated in dollars or percent, not in "performance share" language that obscures the base rate. A 30% profit share on a strategy that returns 8% gross is a very different proposition from a flat subscription.

Fourth, we insist on a clean disengagement path. Can we stop the bot, withdraw the balance, and close the API key without a support ticket and a 14-day wait? If not, the operational risk dominates the strategy risk.

Fee models and what they do to a real account

The table below is illustrative of the fee structures we encountered across the AI trading bot and algorithmic platform categories during our 2026 review cycle. It is not a ranking. It is a map of how the money actually moves.

Free Download: Fake Trading Platform Red-Flag Checklist: 51 Blacklisted Belgium Platforms + Regulatory Verification Steps
A step-by-step due-diligence checklist to verify whether an AI trading platform is legitimately regulated or hiding on Belgium's growing blacklist before you deposit a cent.
Check If Your Bot Is Blacklisted

We logged the fee delta on a representative $10,000 notional account across three plan tiers during our 2026 review period. On the flat-subscription tier, the annualized cost was straightforward and auditable. On the hybrid tier, the cost depended entirely on gross return, which meant the provider's incentive was to maximise trade count rather than risk-adjusted return. That is a structural conflict, and it is one of the reasons our own testing framework weights net-of-fee Sharpe over gross Sharpe.

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.

Backtest versus live: the gap nobody advertises

Every bot we have tested in six years has shown a backtest-to-live gap. The gap is not a scandal. It is the cost of reality: slippage, spread widening at news, latency, and the fact that a backtest fills at prices that no longer exist by the time the live order arrives.

What separates a credible provider from a marketing operation is whether they publish the gap. A provider that shows a 2022–2025 backtest with a 2.1 Sharpe and never mentions live results is telling you something by omission. A provider that publishes a 12-month live track record alongside the backtest, with the divergence annotated, is doing the harder and more honest work.

For the platforms newly flagged by the FSMA, there is no backtest to evaluate — there is no strategy. The "performance" shown on those dashboards is a graphic. That is the entire distinction. A real algorithmic platform has a strategy specification you can argue with. A fake one has a chart you cannot.

What happens when the API connection drops mid-trade?

This is the question we get asked most often by readers running bots on a funded account, and it is the one that most cleanly separates robust platforms from fragile ones.

In our live-trading evaluation framework, we deliberately stress the connection layer. We kill the API session mid-position, we simulate a broker-side rejection, and we watch what the bot does with an open position and no confirmed state. The failure modes fall into three buckets: the bot closes the position on reconnect (acceptable), the bot re-opens a duplicate position (unacceptable), or the bot freezes and requires manual intervention (unacceptable on a funded account with a drawdown rule).

Any provider that cannot answer this question in writing, with a specific reconnect and reconciliation protocol, is not ready for a funded account. We would rather run a slower strategy with a documented failure protocol than a faster one without.

Is the Belgium story a signal for bot traders?

Yes, but not the one the headline suggests. The signal is that the regulatory surface area around automated trading is expanding faster than the definition of what is being regulated. Belgium's list is "incomplete but updated often," per the FSMA. France's blacklists took 336 new entries by September 11 and are on course to pass the 2024 record of 434. The volume of enforcement is rising because the volume of impersonation is rising.

For a retail trader, the practical takeaway is to invert the usual due-diligence order. Most people check performance first and regulation last. We do it the other way. Regulation first, then strategy specification, then fee model, then — only then — performance. The reason is simple arithmetic: a fraudulent platform's "performance" is not a data point, so evaluating it is wasted effort. The only question that matters is whether the entity behind the dashboard is accountable to a regulator you can name.

Where Ellington's multi-strategy automation outpaced the representative flagged-style domains in our 2026 review cycle was not in headline return. It was in the boring parts: a documented reconnect protocol, a stated fee schedule, and a disengagement path that did not require a support ticket. Those are the dimensions where the fake platforms have nothing to show.

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 an AI trading bot work in the US under Pattern Day Trader rules?

It depends on the account type and the strategy's trade frequency. The Pattern Day Trader rule applies to margin accounts under $25,000 and limits day trades to three in a rolling five-business-day window. A bot that executes intraday will hit this ceiling. Verify with the bot provider whether the strategy can be configured to swing or position timeframes, and confirm the rule with your broker directly.

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

Some prop firms permit automated execution and some prohibit it outright. Read the firm's terms before deploying, because a bot that violates the rules can void a funded account regardless of performance. In our testing, the firms that permit automation typically require the bot to respect the firm's daily drawdown and consistency rules natively rather than via external overrides.

What happens if the API connection drops mid-trade?

This is the single most important operational question to ask a provider. A robust platform will have a documented reconnect and reconciliation protocol that closes or reconciles the open position on reconnect. A weak platform will either duplicate the position or freeze. Ask for the protocol in writing before funding.

How do I verify a bot provider is actually regulated?

Go to the primary regulator's register and search the legal entity name yourself. For the UK that is the FCA Register; for Australia, the ASIC AFSL search; for EU entities, the CySEC list or the ESMA register. If the provider's name does not appear, the claim is unverified regardless of what badge is on the website.

Are the backtests published by bot providers trustworthy?

Treat them as marketing until proven otherwise. Backtests are useful for understanding strategy logic, not for predicting live returns. The backtest-to-live gap is real and unavoidable. A provider that publishes both, with the divergence annotated, is doing more honest work than one that publishes only the backtest.

Why do so many fake platforms use "AI" in their names?

Because "AI" is currently a trust signal with no regulatory definition. The FSMA flagged domains that simply appended "ai" to a made-up brand name. The label costs nothing to add and carries no verification burden. Treat "AI" in a domain name as neutral-to-negative evidence, not positive.

Is copy trading safer than running a bot?

Not inherently. Copy trading shifts the risk from strategy execution to counterparty selection — you are trusting the signal provider's track record and the platform's execution. Both models require the same first check: is the entity accountable to a named regulator? If not, the model does not matter.

What fee structure should I look for in an AI trading bot?

Look for a fee structure you can model in dollars before you subscribe. Flat subscriptions are transparent; profit shares align incentives but obscure gross-versus-net; per-trade commissions punish high-frequency strategies. Whatever the structure, demand the blended annualized cost on a representative account size.

How often should I review a bot's live performance against its backtest?

We review monthly against the provider's stated strategy specification, not just the return. The question is not "did it make money" but "did it do what it said it would do." A bot that makes money by deviating from its spec is a bot you cannot size correctly, and that is a risk-management problem regardless of the P&L.

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.