How the channel uses this term ## How the channel uses this term In the channel's conversation with Stripe's Head of Data and AI, Emily Glassberg Sands, multi-account abuse appears as the concrete fraud pattern behind a larger claim: agents are a new class of internet actor, so fraud itself is shifting — from stealing card numbers to targeting compute. [^90e39a1e-87e8-5b3f-b9f2-f003e83fd62b@96] The mechanism is simple: free trials are a generous on-ramp for AI products, so abusers spin up many accounts to drain them. One large AI company observed only 4% of free trials converting to paying customers, with each trial costing $25 in LLM spend — meaning real payers effectively subsidized roughly $625 each, and most trial users were abusers. [^90e39a1e-87e8-5b3f-b9f2-f003e83fd62b@240] [^90e39a1e-87e8-5b3f-b9f2-f003e83fd62b@374] Because the abuse starts at sign-up, not at checkout, defenses are moving up-funnel: fraud is treated as a full-funnel, customer-lifecycle problem integrated at account creation. [^90e39a1e-87e8-5b3f-b9f2-f003e83fd62b@619] - Multi-account abuse = one abuser behind many accounts, exploiting free trials and free credits. - It illustrates why fraud now targets compute rather than card numbers. - The economics are severe: a 4% trial-conversion rate put the cost at about $625 per paying customer in one example. - The countermeasure is up-funnel defense at sign-up, not just payment checks at checkout.