Press Release

Prop Firms Say They Allow AI Trading. Their Rulebooks Say Something Else

Every proprietary trading firm in 2026 has a page that says it welcomes automation. Expert Advisors permitted. Algorithmic strategies supported. Bring your bot.

Then you open the rulebook.

Somewhere between the marketing page and the terms of service, “we allow automation” turns into a list of conditions that quietly excludes most of what a modern trading algorithm actually does. Not through a ban, which would at least be honest, but through a set of narrow carve-outs that are easy to miss until your account is closed and your fee is gone.

This is not a conspiracy. There are real reasons firms write these rules, and some of those reasons are good ones. But if you are a developer or a quant who has spent months building a system and you are about to hand a few hundred dollars to a firm that “supports algorithmic trading,” you should understand what that phrase does and does not mean.

We maintain a comparison database of roughly 60 proprietary trading firms and their published rules. What follows is what those rules actually say.

The gap between “EAs allowed” and “your bot can run”

Start with the most common pattern in the industry: the firm permits Expert Advisors, but only in a supporting role.

For Traders states it plainly. Expert Advisors are allowed as decision-support tools, meaning alerts, partial position management, and predefined rules. Fully autonomous trading bots are prohibited. Read that twice, because it inverts what most people assume. The EA may watch the market and tell you things. It may not decide.

FundingPips draws a similar line with a twist. Third-party EAs are permitted only when used strictly as a trade or risk manager. Any other use of a third-party EA results in denial of the evaluation and closure of the account. There is an exception: an EA you wrote yourself may run full automation. The distinction is authorship, not behaviour, which tells you something about what the rule is really policing. Firms are less worried about automation than about hundreds of traders running the same purchased robot against them.

FundedNext takes the commercial route. Expert Advisors are allowed, but they require a paid add-on. The capability is not a feature of the product. It is a line item.

DNA Funded permits EAs and algorithmic trading on its 1-Phase, 2-Phase and Rapid challenges, and prohibits them entirely on Instant Funding accounts. Same firm, same trader, different product, opposite answer.

Maven does not allow Expert Advisors at all, and only permits copying trades manually.

Five firms, five different meanings of the same sentence.

The rules that were written for high-frequency trading and caught AI instead

Vizuális kereséssel keresett kép

Here is where it gets more interesting for anyone building with machine learning.

Look at what firms prohibit across the board, and a pattern emerges. High-frequency trading. Latency arbitrage. Tick scalping. Statistical arbitrage. Server spamming. Toxic order flow. These appear on nearly every prohibited-strategies list in the industry, including at firms that advertise themselves as automation friendly.

Those are not arbitrary bans. They are the categories where a fast, well-engineered system has a genuine structural edge over the firm’s pricing, and the firm knows it. A prop firm operating on simulated capital makes money when traders fail and loses money when a trader systematically exploits its execution. A model that extracts value from microsecond price discrepancies is doing exactly that.

The problem is that the categories are broad, the definitions are vague, and the enforcement is retrospective. “Tick scalping” is not a precise term. Neither is “toxic trading order flow.” You find out where the boundary sits when someone tells you that you crossed it.

Blue Guardian is one of the few firms that converts this into a number instead of a vibe. Its rule sets a minimum holding time of two minutes. Any trade closed in under two minutes is flagged as tick scalping and counts as a rule violation, and the rule states explicitly that this applies whether the position was closed manually or by an automated system. Average trade duration is expected to stay above one to two minutes.

That is a hard constraint, and for a large class of strategies it is disqualifying. But notice what it does for the trader. It is testable. You can run your backtest, look at your holding-period distribution, and know before you pay whether you comply. Compare that with a firm that reserves the right to decide after the fact that your order flow was toxic.

The most useful rule is not the most permissive one. It is the one you can measure yourself against.

Copy trading is where multi-account automation dies

If your plan involves running one model across several funded accounts, this section is the one that will cost you money.

Almost every firm now restricts copy trading, and the restrictions follow a consistent logic: you may replicate your own decisions across your own accounts, and you may not replicate anyone else’s.

Think Capital allows copy trading across multiple Think Capital accounts, and prohibits copying from personal accounts or from accounts held at other proprietary firms. It adds a condition that matters if you are running a portfolio across several firms: if you want to copy trades between different prop firm accounts, the Think Capital account must be configured as the master. It cannot be a slave account downstream of someone else’s signal.

Blue Guardian permits manual copy trading only between accounts legally owned by the same person, and prohibits copying trades from other traders.

For Traders splits it by direction. Outbound copying, from your For Traders account to external accounts, is allowed. Inbound copying from external sources is strictly forbidden.

FundingPips permits copying between your own accounts and permits your FundingPips account to act as master for external accounts.

The shape is identical everywhere: you are allowed to be the source of a signal, never the destination. Which means the classic scaling move, one model driving ten funded accounts across five firms, is not a scaling strategy. At most firms it is a terminable offence, and the detection is trivial because identical fills across accounts are the easiest pattern in the world to spot.

News, and the four minutes that break event-driven models

If your model reads macroeconomic releases, the calendar is a minefield.

DNA Funded prohibits opening orders within five minutes before or after high-impact news, and extends the rule to unexpected market shocks.

Maven prohibits opening or closing trades within two minutes on either side of a red-folder release, and specifies that the window includes trades hitting take-profit, hitting the profit target, and pending orders. Your model does not need to act during the window to violate the rule. A resting order filling on its own is enough.

FundedNext takes a softer approach that is arguably worse to model around: news trades are allowed, but only forty percent of the profit from trades opened or closed within five minutes of a listed high-impact event counts toward your balance. Your strategy is not banned. Its edge is simply taxed by sixty percent.

Think Capital permits news trading by default only on its Dual Step Swing product. On Lightning, Nexus and Dual Step Intraday it requires an add-on, and the underlying rule blocks all execution, including limit and stop orders, in a four-minute window around high-impact events.

Blue Guardian allows news trading on Challenge, Guardian X and Instant Starter accounts, and disallows it on Instant Standard and funded accounts.

An event-driven model is not a strategy you can port between firms. It is a strategy that has to be rewritten per firm, per product, and sometimes per phase.

Two firms that publish limits instead of hiding them

None of the firms in our database allow unrestricted automation. That firm does not exist, and any firm claiming otherwise has simply moved the restriction somewhere you have not read yet.

The distinction worth caring about is whether the constraint is published as a number or held in reserve as a judgement call. On that axis, two firms stand out.

Blue Guardian permits Expert Advisors without requiring a paid add-on and without carving out an exception for third-party code. Hedging is allowed within a single account. Weekend holding is allowed. Its two-minute minimum holding time is a real constraint that rules out high-frequency approaches, and it is stated as a measurable threshold rather than a vague prohibition on toxic flow. It also publishes exactly which products carry a consistency rule and which do not, which matters because a consistency rule interacts badly with any system whose returns are lumpy. If your model holds positions for minutes rather than seconds, you can verify your compliance before you spend anything.

Think Capital permits Expert Advisors, allows hedging, allows scalping, imposes no consistency rule, and supports MetaTrader 5, ThinkTrader and TradingView. Two constraints deserve to be flagged rather than buried. It caps you at two simultaneous open positions, which is a hard ceiling on any portfolio-style system that wants breadth. And it does not permit VPN use, which is a genuine operational problem if your instinct is to run your model on a remote server and tunnel in. Read that rule carefully against your own infrastructure before you commit.

Neither firm is unrestricted. Both are legible. In a market where the most common failure mode is discovering the rule at the moment it is used against you, legibility is the feature.

What to check before you pay

If you are evaluating a firm for an automated or AI-driven strategy, the marketing page is not evidence. Six questions produce better information:

Does the firm allow autonomous execution, or only decision support? These are opposite answers to the same question, and firms use the same words for both.

Does the permission depend on who wrote the code? A third-party model and a model you wrote yourself are treated differently at several firms.

Is there a minimum holding time? If your median hold is under two minutes, a large part of the market is closed to you regardless of what the automation page says.

What is the copy-trading direction? You can almost always be the master. You can almost never be the slave.

Does the permission survive the phase change? Rules that apply during evaluation frequently tighten on funded accounts, and the strategy that passed is not always the strategy you are allowed to run.

Is the restriction a number or an adjective? “No trades under two minutes” is a specification. “No toxic order flow” is a discretionary power.

The firms worth your capital are not the ones that promise the fewest rules. They are the ones whose rules you can compile against.

Author

  • I am Erika Balla, a technology journalist and content specialist with over 5 years of experience covering advancements in AI, software development, and digital innovation. With a foundation in graphic design and a strong focus on research-driven writing, I create accurate, accessible, and engaging articles that break down complex technical concepts and highlight their real-world impact.

    View all posts

Related Articles

Back to top button