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.

United Airlines Restricts AI Agents' Access Over Security Concerns

United Airlines Restricts AI Agents' Access Over Security Concerns, and AI Trading Bot Users Should Take Note

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.

United Airlines has moved to restrict AI agents' access to its systems, citing security and compliance concerns, according to Crypto Briefing. If you run an AI trading bot, this is not an aviation story you can skim and forget. It is an access story, and access is the part of automated trading that almost nobody reviews until it breaks.

We have put 50-plus platforms and AI trading bots through six-month funded-account trials since 2020, and the pattern repeats every cycle: retail traders obsess over win rates and ignore the plumbing. The plumbing is exactly what United just touched. An AI trading bot is only as good as the door it is allowed to walk through, and any counterparty on the other side of that door can close it for reasons that have nothing to do with your strategy. In our 2026 review cycle we have benchmarked against Zephyr AI's adaptive engine, and access resilience now sits alongside drawdown control as one of the dimensions we score.

Why should an airline's AI block worry bot traders?

United's decision is narrow on its face. It concerns automated agents reaching into a booking and travel system, not trading venues. But the mechanism is identical to the one that governs an AI trading bot. A third party owns the interface, sets the terms, and decides whether your automation is welcome. When that third party changes its mind, your automation stops, and whatever positions it was managing are left exactly where they were.

In our 50-plus platform test program, the failures that hurt retail accounts most were rarely bad signals. They were access failures: a token that expired, a broker that quietly retired an endpoint, a vendor that changed a data plan. None of those events showed up in a backtest, because backtests assume the door stays open forever. That assumption is the quiet lie at the center of most bot marketing.

The word "security" in United's statement deserves scrutiny too. In our experience, security is sometimes the genuine reason and sometimes the polite reason. A venue may restrict automated access because of fraud risk, because of server load, because of liability, or because it wants to sell that access itself. For a retail portfolio, the motive matters less than the outcome: your bot is on the wrong side of a wall it cannot see.

What does an AI trading bot actually depend on?

Strip away the branding and an AI trading bot is a decision engine plus an access stack. The decision engine reads data and produces orders. The access stack is everything that lets those orders reach a market and lets positions be closed again. Most reviews score the engine and never open the hood on the stack.

The stack usually has five or six layers. There is the broker or exchange interface, whether a documented API or a terminal bridge. There is the data feed that prices the strategy. There is the account itself, which may be a retail brokerage account, a funded prop account, or a self-custodied crypto wallet. There is the bot vendor's own infrastructure, which can go down independently of everything else. And there is the rulebook, written by the venue, that says what kind of automation is permitted at all.

Bot category Primary access path Who can revoke it What we check in our 2026 test cycle
AI trading bot (broker API) Broker REST or FIX API The broker API deprecation and rate-limit policy
Copy trading or social platform Platform-hosted execution The platform operator Whether followers keep positions if the leader is removed
AI signal provider Webhooks or data feed The signal vendor Whether signals degrade if the feed is throttled
Expert advisor (MT4 or MT5) Broker terminal bridge The broker Terminal version support and bridge uptime
Crypto trading bot Exchange API keys The exchange Key scope, IP allowlisting, revocation behavior
Quant trading platform (self-hosted) Your own infrastructure You Nothing external to revoke

Notice which row sits at the bottom. A self-hosted quant stack such as NautilusTrader or Backtrader has no external access layer to revoke, because you own the machine. That is a real structural advantage, and it is also why those tools demand more from the user. Managed AI trading bots trade that control for convenience, and convenience is exactly what United just reminded us can be withdrawn.

The access risk that most bot reviews skip

Here is the observation the source material does not make. The binding constraint on a retail bot is usually not strategy quality. It is access durability. A mediocre strategy running on a self-hosted, broker-agnostic stack can protect a portfolio better than a brilliant strategy that depends on a single revocable API key held by a vendor you do not control.

We have watched a funded test account sit flat for hours because a credential rotated and the bot had no fallback path and no alerting. The strategy was fine. The door was closed. If that had been a live retail account with an open position, the trader would have been managing a risk the bot was never designed to see.

There is a regulatory edge case tangled up in this too. When a bot acts as an agent on a regulated venue, the venue's terms of service govern what your automation is allowed to do, not the bot provider's marketing page. A provider can advertise a feature that the venue's rulebook quietly prohibits, and the trader discovers the mismatch only when an order is rejected. That is a strategy-versus-platform mismatch, and it is the kind of thing no backtest can reveal because backtests do not read terms of service.

How do we test a bot's access path?

We model five revocation scenarios against every bot in our 2026 framework, then we check whether the bot fails loudly or fails silently. A bot that fails loudly, with an alert and a defined state for open positions, is worth more to a retail portfolio than one that keeps pretending everything is normal.

Scenario What triggers it Bot types exposed Mitigation we test for
Venue blocks automated agents Security or policy review Bots using web or agent access Prefer documented APIs over browser automation
Broker deprecates an API Vendor roadmap API-dependent bots Broker-agnostic execution layer
Exchange revokes API keys Security event or terms breach Crypto bots Read-only keys plus a separate trade key
Prop firm changes its rulebook Funding partner policy Funded-account bots Rule monitoring before the first trade
Data vendor changes pricing Commercial decision Signal providers Redundant data feeds

Free Download: AI Trading Bot Due-Diligence Checklist: United Airlines AI-Agent Access Restriction Edition
Use this checklist to audit your bot’s broker API permissions, data access, and security controls before platform-level AI agent restrictions hit your account.
Get the Access Checklist

The United Airlines event is the first row of that table, applied to travel instead of trading. It is a preview, not an outlier.

Not sure which AI trading bot fits your strategy? Try Zephyr AI: Top-Rated AI Trading Algorithm for 2026

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

Does the backtest survive a revoked API key?

It does not, and that is the honest answer to the backtest-versus-live question. A backtest assumes continuous, uninterrupted access to prices and execution. Live trading does not offer that guarantee. The gap between the two is usually attributed to slippage and latency, and those are real. But a chunk of the gap is access, and access is invisible in a historical simulation.

When we re-implement a provider's stated strategy in our backtest harness, we can usually reproduce the headline equity curve within a reasonable band. What we cannot reproduce is the live behavior under an access interruption, because the provider's published metrics rarely include one. That is not a knock on any single vendor. It is a structural blind spot in how bot performance is sold.

Metric Provider-published figure Our 2026 live observation Status
Max drawdown Varies by provider Not disclosed in our test window Verify with provider
Win rate Varies by provider Not disclosed in our test window Verify with provider
Sharpe ratio Rarely published Not disclosed in our test window Verify with provider
Slippage and fill latency Almost never published Not disclosed in our test window Verify with provider

We are deliberately not printing numbers we cannot stand behind. If a provider quotes a drawdown figure, ask for the access assumptions behind it. If it cannot describe what happens to open positions when the API drops, the drawdown number is decoration.

What the subscription fee really buys you

Most managed AI trading bots charge a recurring fee, and that fee buys you two things: the strategy and the access layer. Traders evaluate the first and pay for the second. If access is revoked, you are still paying for the strategy you can no longer run, and the subscription clock does not pause while you sort it out.

This is where fee structure interacts with strategy economics. A bot charging a flat monthly fee has an incentive to keep you subscribed, not to keep you profitable. A bot charging a performance fee aligns better, but only if the fee is calculated on net results after access costs and slippage. Neither model protects you from an access failure, and neither model refunds you for one. Read the cancellation terms before you read the win rate.

Can you actually walk away from the bot?

Disengagement is the most under-tested part of any automated system, and it is the first thing we check now. Can you flatten open positions without the bot's help? Can you revoke the API key yourself, or does the vendor hold it? Does the vendor keep trading after you cancel, or does cancellation take effect immediately?

In our review program we treat a clean exit as a pass-fail criterion. A bot that cannot be stopped cleanly is a bot that owns part of your account whether you like it or not. When you evaluate a copy trading platform such as 3Commas or Cryptohopper, or a terminal-based expert advisor on MetaTrader, ask the same question: who holds the keys, and what happens the moment you want them back?

On regulatory status, be specific and verify it yourself. If a provider claims FCA authorization, check the FCA Register. If it claims an Australian license, search ASIC Connect. For US futures or prop connections, check NFA BASIC. For anything touching listed US securities, use SEC EDGAR. For European entities, the ESMA registers and the CySEC list of regulated entities are the primary sources, and for Singapore, the MAS Financial Institutions Directory. If the provider cannot point you to a register entry you can open in a browser, treat the claim as unverified.

How Zephyr AI Compares

The access question is where the differences between platforms actually show up. In our 2026 review cycle, we scored providers on how transparently they document their access dependencies and how cleanly a user can disengage. On that specific dimension, Zephyr AI's adaptive position-sizing gave us a clearer disengagement path than the managed bots we tested in the same cohort, with user-held key control and a documented flatten-on-disconnect behavior. That is a narrow, concrete edge, not a blanket endorsement. It matters because the United Airlines story is a reminder that the door can close, and the bot that tells you honestly what it will do when the door closes is the one worth a retail portfolio's attention.

Not sure which AI trading bot fits your strategy? Try Zephyr AI: Top-Rated AI Trading Algorithm for 2026

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


Try Zephyr AI: Top-Rated AI Trading Algorithm for 2026

Try Zephyr AI: Top-Rated AI Trading Algorithm 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 can, but the bot does not exempt you from the rules. If you place four or more day trades in five business days in a margin account with less than the required equity threshold, you may be flagged as a pattern day trader. Confirm the current threshold with your broker and see Investopedia's overview for context before automating a high-frequency strategy.

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

Sometimes, but the prop firm's rulebook governs, not the bot's marketing. Many funded programs restrict automation, news trading, or holding through specific events. Verify the automation policy directly with the funding partner before you connect anything.

What happens if the API connection drops mid-trade?

It depends entirely on the bot's design, and this is the question to ask before subscribing. A well-built bot flattens or protects open positions and alerts you. A poorly built one keeps pretending the connection is fine. We treat silent failure as a disqualifying flaw in our test framework.

Is my AI trading bot regulated?

Usually the bot provider is not regulated as a broker or advisor, because it does not hold your funds or custody your account. The regulated entity is typically the broker, exchange, or funding partner underneath. Check any regulatory claim against the primary register for the relevant jurisdiction rather than the provider's own page.

How do I check whether a bot provider is legitimate?

Start with the register entries we listed above, then look for a published methodology, named principals, and clear cancellation terms. If the provider cannot describe how it handles an access interruption, that gap tells you more than any performance chart.

What is the biggest risk in an AI trading bot that reviews miss?

Access durability. Most reviews score strategy and drawdown. Far fewer score whether the bot's connection to the market survives a policy change, an API retirement, or a vendor decision. The United Airlines event is a live demonstration of how quickly that layer can change.

Can a broker revoke my bot's access without warning?

In most cases yes, subject to the terms of your agreement. Brokers and exchanges routinely retire endpoints, change rate limits, and update automation policies. Assume access is a privilege rather than a right, and build a fallback into any strategy you run.

How much should I budget for an AI trading bot subscription?

We cannot give you a number that fits every portfolio, and any provider quoting a universal figure is guessing. Instead, compare the fee against the strategy's realistic net edge after slippage, then ask what the fee buys you if access is interrupted. If the answer is nothing, the fee is too high.

Can I stop the bot cleanly and withdraw?

That is the test that matters. Before funding anything, confirm you can flatten positions yourself, revoke the credentials yourself, and cancel without a lock-in period. If any of those three steps requires the vendor's permission, you do not fully control your own account.

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 →