How I fight token fraud as an AI founder
Claire, a solo founder, argues that token fraud is a major, costly threat to AI products, and that selling AI means selling compute, so free doors must close and controls must layer across business, code and agents.
Watch on YouTube ↗Charlie's summary
- ▶ 2:01
Token fraud has cost a bootstrapped founder tens of thousands of dollars, and even big companies face it
Claire runs ChatPRD on seat-based pricing built around average usage. People who are not customers can still abuse free trials, promo codes and early access to run up her LLM bill. It has cost her “tens of thousands of dollars” over the past year. When she posted about it on X, founders from large companies and strong startups replied that they deal with it daily, and one friend was about to turn off signups.
- ▶ 5:03
Six abuse patterns: trial farming, promo codes, bad cards, relays, MCP workers and model probing
Claire lists the ways people abuse ChatPRD. They farm free tiers with the most expensive models, resell promo codes, use prepaid cards and chargebacks, and resell access through relays and reverse proxies. One person installed a custom MCP that ran as a 24/7 worker. Others probed which model it was. She says the worst abusers are commercial criminal enterprises, and trial abuse from low-income countries is the least nefarious. Her examples include a Roblox and Fortnite map studio running loops of trials and promo codes.
- ▶ 6:06
Free trials are over: every protection against fraud costs real business, so close the zero-dollar doors
Claire turned off free trials, her best source of PLG and enterprise leads, because she could not build a cost-effective free tier that wouldn't be abused. Credit-card-required and paid trials failed too, since fraudsters cycled bad cards and charged back. She is also dropping promo codes. Her view is that you must accept the pipeline hit and move to sales-led access. Others in the industry are doing the same.
- ▶ 14:49
Business controls: bill up front, set hard limits, keep kill switches, and consider blocking countries
Following a list from Gareth at Atlassian, Claire recommends charging up front with no grace period on failed payments. She also advises hard limits per account and organization, restricting trials to business email domains, feature flags to switch trials off, and terms of service that allow banning suspected abuse. Turning off whole countries via firewall made fraud drop significantly, though it hurts legitimate users. She also filed cyber crime reports with the FBI and urges founders to share fraud patterns openly.
Give your email to unlock the rest of the key points — for every episode.
Do more with Charlie
Free — sign in with your email, no password.