Guide

Free trial abuse and credit farming on AI platforms

When every free account comes with GPU time, model tokens, sandboxes or browser sessions, one person opening twenty accounts is no longer a rounding error. Here's how it happens and what you can do about it.

What is credit farming?

Credit farming is when one person or group creates many accounts to collect the free credits each new signup receives. On a traditional SaaS product a duplicate account mostly costs a database row. On an AI or infrastructure product it costs real compute bills: model calls, GPU minutes, microVMs, proxy bandwidth or phone minutes.

Common tactics

  • Email aliases: name+1@gmail.com, dotted Gmail addresses and disposable inboxes that all land with the same person.
  • Same device, new identity: a fresh email and name from the same browser or machine.
  • Rotating networks: VPNs, residential proxies or virtual phone numbers to look like different people.
  • Automation: scripts and headless browsers that sign up and spend credits without a human.

Why the usual fixes hurt growth

Requiring a credit card, phone verification or CAPTCHAs on every signup adds friction for genuine developers too. Blocking a signup outright is risky because a single shared signal, like an office network, can belong to many real people.

A different approach: share one limit across linked accounts

Halo connects signals at signup — normalized emails, devices, networks and phones — into linked accounts. Then, after signup, it checks each costly request against a daily spending limit shared across those linked accounts. Creating another account doesn't mean getting another full allowance.

Halo recommends allowing a request, slowing it down, using a cheaper model or blocking it. Your app applies the decision, so you stay in control.

Start by measuring, not blocking

The safest first step is shadow mode: Halo scores signups for 72 hours without changing anything for your users, then produces a Leak Report showing how much free credit was flagged as going to suspected duplicate accounts. You decide whether to turn on enforcement once you've seen the numbers.

Checklist

  • Know what one free signup costs you in compute.
  • Normalize emails before comparing them.
  • Look at several signals together, never one alone.
  • Limit spending across linked accounts instead of only gating signup.
  • Keep a human review path for borderline cases.

See your own numbers in 72 hours

Add one script tag in shadow mode. Halo blocks nobody, then sends a Leak Report showing how much free credit went to suspected duplicate accounts.