Short answer: Yes. Prop firms use automated risk software to watch every trade for martingale sizing, high-frequency patterns, copy trading, hedging between accounts and news-window breaches. Some of it is marketed as AI; much of it is still rule-based scripts with a new label. It flags patterns, not intent, so a legitimate strategy that looks like an abusive one can be flagged, and the burden of proof usually lands on you. Know what trips the alarms, keep your own records, and question any breach notice in writing.
- February 2026: Quant Technology Group announced general availability of QuantSentry, a risk platform that claims to flag copy trading, hedging schemes and multi-accounting before payouts are released
- 5 to 10: the size of the coordinated account cluster that one industry strategist says can pull out meaningful payouts before internal alerts fire
- 42% to 32%: the payout-to-revenue drop one vendor advertises for a “typical” $100,000 challenge after automated profit deductions (a vendor claim, with no published method)
On this page
- Algorithm detection questions, answered
- What the software is actually watching
- Where “AI” is real and where it is a rebrand
- The five patterns that get honest traders flagged
- How firms use the output: warnings, deductions and breaches
- How to protect your account
- How to question a flag
- Verdict
- Frequently asked questions
Algorithm detection questions, answered
Do prop firms really use AI to monitor traders?
Yes, in the sense that automated systems score your behaviour continuously. Vendors such as Quant Technology Group (QuantSentry) and Arizet (PropRisk) sell platforms that monitor accounts in real time and apply machine-learning or network analysis on top of fixed rules. Whether a given firm uses a full model or a set of if-then scripts is rarely disclosed.
What do these systems look for?
Strategy patterns (martingale, grid, high-frequency), coordination between accounts (copy trading, mirrored or opposite positions), rule breaches (lot size, news windows, exposure caps) and identity signals such as shared devices or IP details.
Can an algorithm fail a trader unfairly?
It can flag one unfairly. A flag is not a verdict, but depending on the firm it can trigger a review, a profit deduction or an account breach, and you often cannot tell how closely a human looked. That gap is where traders get hurt.
Is this the same as a firm changing the rules on you?
No. Detection enforces rules that already exist, often buried in the terms. A rule change is a different problem, and we cover it in the Retroactive Rule Change Playbook.
What the software is actually watching
Risk monitoring in a prop firm is built around one question: is this account generating profit the firm can safely pay out? Everything else is a way of answering it.
Most platforms track four layers at once.
Trade behaviour. Position sizing after losses, trade frequency, holding times, and how risk changes as the account approaches a target or a drawdown limit. This is where martingale and grid patterns show up.
Rule compliance. Maximum lot size, daily and overall drawdown, consistency rules, and restrictions around news releases. These are the easy checks, because they are binary.
Cross-account relationships. Do several accounts enter and exit the same instrument in the same tight window? Does one account only profit when another loses? This is the layer that targets copy trading and hedging.
Identity and device signals. Browser, operating system and hardware fingerprints, and IP or VPN use. One 2026 industry explainer notes that these signatures can persist across network changes. Some firms restrict VPNs and remote tools, and the rules differ, so check yours.
Nobody publishes exact thresholds. A copier-software explainer we reviewed says no firm discloses its fill-timing window, and treats the millisecond figures floating around forums as unverified. We agree with that caution. If a Discord channel tells you “the limit is exactly 3 IPs a day” or “anything under 10 ms gets flagged”, treat it as rumour until a firm’s own rulebook says it.
Where “AI” is real and where it is a rebrand
This is the part the marketing blurs.
On the vendor side, the claims are getting bolder. Quant Technology Group’s February 2026 launch announcement says QuantSentry uses network analysis and machine learning to detect coordinated abuse such as copy trading, hedging schemes and multi-accounting, and to catch abusive behaviour before fraudulent payouts go out. It also says the platform produces PDF “evidence kits” that firms can use to enforce their terms. Arizet’s PropRisk page lists detection of HFT, martingale, copy trading and grid trading, and automated profit deductions after violations.
Two things are worth noticing.
First, none of this is independently tested in anything we could find. The announcement contains no customer results or performance data, and PropRisk’s headline case study (payout-to-revenue falling from 42% to 32% on a typical $100,000 challenge) comes with no sample size or method. These are sales claims. They tell you what firms are being sold, not how accurate it is.
Second, the real technical shift is the move from single-account checks to network-level analysis. A May 2026 Finance Magnates piece by Arizet Labs’ chief strategy officer describes it plainly: old rule-based systems catch identical entries and uniform sizing, but struggle once coordinated accounts spread across brokers, devices and slightly varied timing. The article says the newer approach maps behaviour across accounts as a graph, that it is still early, and that no industry-wide detection layer exists. It also stresses that correlation alone is not proof of copying; persistent alignment across timing, direction, size and co-movement raises the probability, but does not establish intent.
That last point matters for you. The vendors themselves concede the evidence is probabilistic.
| Approach | What it catches | What it misses or misfires on |
|---|---|---|
| Fixed rules (lot caps, drawdown, news windows) | Clear, binary breaches | Nothing subtle; also punishes harmless edge cases |
| Pattern flags (martingale, grid, HFT) | Obvious risk-escalation behaviour | Legitimate scaling-in or scalping that looks similar |
| Cross-account correlation | Copy trading and mirrored hedges | Coincidence between unrelated traders on the same setup |
| Network or ML scoring | Coordinated clusters that vary timing | Opaque decisions you cannot easily check |
The five patterns that get honest traders flagged
Most traders who get flagged are not running an abuse scheme. They are doing something that resembles one. These are five patterns that commonly trigger flags.
1. Scaling in that looks like martingale. Adding to a position after a loss can be a deliberate plan with a fixed risk budget. To a script that watches size after drawdown, it can look identical to doubling down. If your strategy involves adding to losers, check the firm’s definition of martingale and grid before you trade.
2. Fast scalping that looks like HFT. Many scalpers hold positions for seconds or minutes. Firms define high-frequency in different ways, and some set minimum holding times or trade-count ceilings. Read the exact wording rather than assuming “scalping allowed” means “any speed allowed”.
3. Trading the same idea as thousands of others. If a popular signal or indicator drives your entries, your timing can match strangers’ accounts at the same firm. Correlation software is looking for persistent alignment, not a one-off coincidence, but a widely followed setup raises the odds of a false match.
4. Holding through a news window. Enforcement here depends on timing, not profit. Firms that restrict news trading set windows of a few minutes around major releases such as NFP, CPI and FOMC, and holding a winning position through one can still be a breach. Some firms apply the restriction only to funded accounts, and windows change, so confirm the current rule in the firm’s own terms. We track which firms allow news trading in Prop Firms That Allow News Trading.
5. Running several of your own accounts. Account limits and capital caps are enforced automatically, and mirrored lot sizes across your own accounts add up to one large position in the firm’s eyes. As of 22 October 2026: FTMO’s own help page sets a $400,000 simulated-capital cap per trader or strategy and says identical or substantially similar trading across accounts, including copied signals, can lead to suspension if the cap is exceeded. FundedNext’s help centre caps linked accounts at $300,000, allows copying only between your own challenge accounts, and says a first violation brings a warning and no Performance Reward for that cycle, with termination after a repeat. Topstep’s help centre allows five active Express Funded Accounts at once. Apex’s help centre (last modified 15 April 2026) sets a limit of 20 active Performance Accounts across everyone in the same household, companies and connections, and allows copy trading across them.
One rule is common: opposing positions across accounts are treated as hedging. Topstep’s help centre prohibits it explicitly, and Apex requires all Performance Accounts to trade the same direction and not hedge against correlated assets. Permission to copy trade does not override it.
How firms use the output: warnings, deductions and breaches
A flag can end in very different places.
- A warning or request for information. The best case. You explain the strategy and carry on.
- A profit deduction. PropRisk advertises automatic deduction of profits the system judges illegitimate, with manual adjustment available. If your payout arrives lower than expected, ask which trades were removed and under which clause.
- A breach or account closure. The worst case, and the one that most often ends in a dispute.
- A refused payout. Some firms run their checks at the payout stage. QuantSentry is marketed on catching abuse before payouts are released, which means a flag can surface after weeks of trading.
The structural problem is incentive. A tool whose headline benefit is lower payout leakage is not neutral about the outcome of a borderline case. That does not make firms dishonest. It means you should assume the burden of proof will land on you and prepare accordingly.
If you are on the receiving end of a changed or disputed rule, the Retroactive Rule Change Playbook walks through the first 72 hours, and our Trust Index scores firms on rule stability and payout record. For the wider context on why firms are tightening controls, see the 2024 to 2026 prop firm collapse.
How to protect your account
Do these before the first trade, not after the first warning.
- Read the exact clauses on martingale, grid, HFT, copy trading, hedging and news trading. Quote them into a document with the date and a link. Rulebooks get edited.
- Screenshot or archive the rules page on the day you buy. The Wayback Machine works if the firm has not blocked it.
- Keep a trade journal that records your reason for each entry. A written strategy rationale is the best defence against a pattern flag.
- Use one identity, one device setup and a stable connection. Avoid VPNs unless the firm explicitly allows them.
- Do not share signals or copy another trader’s account. Follow a public idea if you like, but size it and time it yourself.
- Stay flat around the firm’s news windows if you are not sure. The cost of waiting is small compared with a breach.
- Ask support in writing before you do anything unusual. Save the reply.
- Export your trade history regularly. If the dashboard disappears, you still have the record.
How to question a flag
If you get a warning, deduction or breach notice, keep your reply short and factual.
Hello, I received a notice on [date] regarding account [number]. Please tell me: (1) which specific rule was breached, (2) which trades or time period triggered it, (3) whether the decision was made by an automated system or reviewed by a person, and (4) how I can appeal. I have attached my trade journal and the rulebook version I accepted on [date]. Thank you.
Three things to watch for in the answer. A vague reference to “risk policy” with no clause is a red flag. A refusal to name the trades involved is another. And a firm that cannot say whether a human reviewed the decision is telling you something about how the process works.
Verdict
The headline of this article is deliberately blunt, and it is only half fair. Firms use automated detection mainly to protect their capital, and some of the abuse they target (coordinated account clusters, mirrored hedges, multi-accounting) is real. An industry strategist writing in Finance Magnates says undetected clusters can inflate payout ratios and distort who is actually skilled.
But the technology is sold with confident numbers and no public accuracy data, the thresholds are secret, and the vendors themselves say correlation is not proof of intent. Put those together and you get a system that can wrongly flag an honest trader, with the cost falling on the trader.
Our view: you cannot beat the software, and you do not need to. Trade a strategy you can explain, read the rules that apply to it, keep your own records, and choose firms that publish clear definitions and an appeals route. If a firm cannot tell you why you were flagged, that is information about the firm.
Sources checked on 22 October 2026. Vendor claims are marketing statements and have not been independently verified. Firm rules, account limits and news windows change often; confirm them in each firm’s current terms. JoinProp is not a law firm or financial adviser, and this article is educational, not legal or investment advice.
Frequently asked questions
Do all prop firms use AI to detect trading behaviour?
Not necessarily. Firms that license third-party risk platforms are the ones most likely to run advanced automated monitoring; others may rely on rule-based alerts and manual review. Firms rarely say which they use, so read the terms and ask support.
Is martingale banned at prop firms?
It depends on the firm. Some restrict or ban it, and definitions differ. Some treat any systematic doubling after a loss as martingale; others only flag it when risk exceeds a set limit. Check the firm’s own definition.
How do prop firms detect copy trading, and is copy trading allowed?
Policies differ: some firms allow copying between your own accounts up to a capital cap, others ban it outright. Detection is typically through timing correlation across accounts, matching position sizes and directions, shared device or IP signatures, and position mapping. Newer tools analyse behaviour across a network of accounts rather than one account at a time. Correlation alone is not proof, and the strength of the evidence varies.
Can a prop firm take my profits because of an algorithm flag?
If its terms allow profit deductions or breach decisions for rule violations, yes, and some risk platforms automate this. Whether it is fair is a separate question. Ask for the clause, the trades involved and the review process.
Is scalping the same as an HFT rule at a prop firm?
Not necessarily. Most firms allow scalping, but an HFT prop firm rule may cap trade counts or set minimum holding times. Read the wording on HFT and holding periods in your firm’s rulebook.
Is hedging allowed at prop firms?
Hedging inside one account is sometimes permitted; opposing positions across accounts, or across firms, are commonly treated as prohibited prop firm hedging. Confirm the wording in your firm’s rulebook, because copy-trading permission does not override it.
What should I do if my account is breached for a pattern I did not intend?
Stop trading the account, export your trade history, save the rules you accepted, and write to support asking for the specific clause and trades. See the Retroactive Rule Change Playbook for the evidence checklist.
Are the detection thresholds published anywhere?
Almost never. Figures quoted in forums, such as exact millisecond windows or IP-per-day limits, are usually unverified. Rely on the firm’s published rules instead.
Which firms are safest if I want clear rules?
Look for published definitions of banned strategies, a stated appeals process and a clean payout record. Our Trust Index scores firms on factors like these.
Sources
- Quant Technology Group, QuantSentry launch announcement (10 February 2026), via FXStreet
- Arizet, PropRisk product page (Match-Trader ecosystem)
- FTMO, FundedNext, Topstep and Apex help centre pages (checked 22 October 2026)
- Shervin Arian, “The Systemic Cost of Copy Trading in Prop Trading Firms,” Finance Magnates (5 May 2026)
- ThorTradeCopier, “How Prop Firms Detect Copy Trading” (May 2026)
Enjoyed this article? Add JoinProp as a preferred source on Google.
