Most AI companies already run Stripe, so the question is what Radar covers and what is left. Radar is strong on payments and, on its Pro plan, now scores trials, sign-ups and logins. It has little to go on once an approved account starts calling your API.
GuideHarry Johnston, NoResponse9 min read
Radar covers payment fraud and, on Pro, selected signup, login and billing signals. It does not inspect model requests, token resale or API-only ban evasion. AI platforms need a decision before provisioning credits and a usage layer after the first request.
This guide answers the question from Stripe's own documentation, pricing page and blog posts, each read on 19 October 2026. For the wider picture of how tokens leak from a platform, start with our token fraud guide.
At a glance
Radar sees payments across Stripe's whole network, so it can stop a stolen card on its first visit to your checkout. For an AI company that sells credits or subscriptions, that is the largest single fraud loss Radar removes.
Stripe's 15 September 2026 post on fraud at AI startups describes the pattern it was built for. A fraudulent actor buys a subscription or a block of tokens with a stolen card and resells it, and "By the time the legitimate cardholder notices and disputes the charge, the resale is complete and the business absorbs the loss." The same post makes the case for network data. "Because fraudulent actors rotate through stolen cards across businesses, this pattern can be difficult for any single company to detect on its own."
Three more strengths matter to AI buyers:
Our card testing and chargebacks guide covers the Radar rules worth turning on for credit top-ups.
An AI company that gives away free credits or trials needs Radar Pro for the abuse controls. Lite, Standard and Plus cover payments, and Plus adds custom rules and manual reviews.
Stripe's documentation names four plans. Lite is "AI-based fraud prevention for payments and card testing, plus fraud alerts." Standard adds protection for all payment methods and default rules. Plus adds custom rules, backtesting, risk scores and manual reviews, the tools Stripe has also sold as Radar for Fraud Teams. Pro is described as "Advanced protection against emerging fraud threats and customer abuse. Detect multi-account, free trial and pay-as-you-go abuse."
Under customer abuse prevention, the comparison table marks the free trial control, bot abuse, the multi-account signal, the account sharing signal and the pay-as-you-go signal as supported on Pro only.
Stripe's Radar pricing page shows a monthly subscription and a pay-as-you-go option for Standard, Plus and Pro, with higher prices for platforms and marketplaces. The prices vary by country, and the page served us local prices rather than US dollars, so check it for your own account. Stripe's documentation adds that these three plans charge a fee for each transaction Radar screens, while Lite comes at no extra charge on standard payments pricing.
The control scores the payment method a customer saves when a trial starts, and blocks the trial if the risk is high. It works only when Stripe can tell which card collections are trial starts, which depends on how you integrate.
Stripe's free trial abuse documentation says Checkout Sessions in subscription mode detect trial starts with no code changes. The Subscriptions API needs a one-time change. Setup Intents, Payment Intents and other paths need you to email Stripe so it can help you add the metadata. You then switch on Radar for saved payment methods and enable the control, and Stripe lets you backtest it on recent trial starts first.
The March post gives Stripe's results for the control. It predicts abusive trial behaviour "with 90% accuracy" and, across four AI businesses in its first two months, blocked more than 550,000 high-risk trials worth an estimated $4.4 million in compute. These are Stripe's own figures, and the post gives no method for either.
The limit follows from the design. The control judges a card at trial start. If your free credits need no card, there is no trial start for it to judge. Stripe's sign-up figures, quoted in several different wordings, are set side by side in our free trial abuse guide.
With Radar Pro, Stripe can now score a sign-up that has no card, through its account evaluation API. That API is new, is in public preview and judges the registration and login events you send it.
Stripe's account evaluations documentation says "To use the API in live mode, you must have Radar Pro or a Radar Pro trial." The changelog of 26 August 2026 lists the Account Evaluations, Account Activity and Account Signals APIs as public preview.
At registration, you create a Radar Session with Stripe.js in the browser, then request a multi-accounting score from your server. Stripe returns a score from 0 to 100 and a risk level, and asks you to report whether you allowed, restricted or blocked the account. Its example of a restricted account is one that "receives fewer promotional credits", which suits an AI company's free credits well.
Three details matter for AI sign-ups:
Radar Pro also scores bots, but only on payments. Its bot abuse documentation says the score applies to payments made through Checkout, Custom Checkout, Radar Sessions or Web Elements.
After the payment, Radar watches logins for account sharing and growing usage bills for non-payment. It does not watch API traffic, so it has little to say about resale, shared API keys or banned users who return.
Account sharing. Stripe defines its account sharing signal as "Risk that the same account is being used from multiple locations simultaneously", evaluated at login. Stripe documents account sharing at login. That catches people sharing a web login. A relay, which resells cut-price access to models it does not own, sends its customers' requests from its own servers to your API with an API key. It never logs in.
Pay-as-you-go abuse. The pay-as-you-go documentation says to "Use this signal for post-paid billing models where customers accumulate usage before being charged." You simulate the next invoice and Stripe returns a non-payment risk level. It answers whether the customer will pay. An account that resells your tokens and pays its bill on time scores as a good customer.
Resale. Stripe names resale as a motive in its September post, and its remedy there is the stolen card check. Stripe's public Radar documentation does not describe a control for an account that pays with a clean card and sells your tokens through a relay. Our token resale guide covers the usage signals that do.
Ban evasion. The multi-account score draws on prior abuse across Stripe's network, according to the May post. Stripe's outcome reports cover registration and login decisions. When you ban an account weeks later for how it used your API, the evidence for that ban lives in your logs, and our ban evasion guide covers how to match a new account to it.
Keep Radar on every payment and switch on the Pro controls if you give free trials or credits. Add a layer that checks each sign-up before credits go out and watches how each account uses your API afterwards.
The two layers answer different questions. Stripe supplies payment and selected signup risk signals; NoResponse adds device intelligence, behavioural fingerprinting and account-link analysis across API usage. Radar asks whether this card, device or login looks like abuse Stripe has seen before. A sign-up and usage layer asks whether this email address or phone number has any history outside your platform, and whether this account's traffic looks like resale or like an account you banned.
The order we recommend:
Each check uses data you already hold in Stripe, your sign-up table or your gateway logs.
We work on the account, and Radar keeps the card. Your server calls our sign-up check with the new account's email address, username or phone number. We search about 4,000 sites for that identity and send back a risk score, the reasons and a decision in about three seconds, before the account's first request. The call needs no browser script, so a sign-up from a command-line tool gets the same check.
We build a usage fingerprint for each account in your gateway from its API traffic. Fingerprints match a new account to one you banned and pick out accounts reselling your tokens, which are the two patterns Radar's login and billing signals leave open. With relay intelligence you see where your tokens are sold among the 70+ relay and resale services we track, and which accounts and API keys feed each one. Incident support puts our team beside yours while an attack is running.
You can report the decision you make on our score to Stripe as an account outcome, so Radar learns from it too. Book a demo and we'll show you which of last month's paying accounts we would have flagged.
Every token an AI platform serves runs on compute someone pays for. Token fraud is when the person using it is not the person you sold it to. This guide maps the four ways that happens and links to a guide for each.
GuideEvery free credit on an AI platform is compute someone pays for. Score each sign-up before it gets credits, stage the credits and keep the card for accounts that raise a signal.
GuideA stolen card that buys API credits costs you twice. The compute is spent as soon as the tokens are served, and the dispute arrives weeks or months later with a fee and a mark against your card network ratios. Here is how the pattern works and where to stop it.
GuideA ban removes an account, not the person or the setup behind it. On an AI platform the strongest evidence that they are back is how the new account uses your API.
GuideA resold account looks like one customer and behaves like a hundred. This guide sets out the signals in your own logs that give it away, how to confirm resale without hurting real customers, and what to do once you find it.
GuideRadar covers payment fraud. It does not inspect model requests, token resale or API-only ban evasion. Use a pre-provisioning decision for signup risk and a usage layer for what happens after the first request.
Radar Pro. Stripe's plan comparison lists the free trial control, the multi-account signal, the account sharing signal, the pay-as-you-go signal and bot abuse as Pro features. Stripe also offers a Radar Pro trial.
Radar can stop resale funded by a stolen card at payment. It does not describe a control for an account that pays with a clean card and resells your tokens through a relay, because that behaviour appears in API traffic.
