Heedeo exposes 500+ repos farming AI credits
Anti-fraud platform Heedeo published an investigation revealing over 500 GitHub repositories actively automating trial abuse across 14 commercial AI products. Attackers systematically spoof hardware fingerprints, rotate burner emails, and proxy requests to siphon expensive GPU inference on founders' tabs.
The freemium SaaS playbook is breaking under AI unit economics, where unauthenticated onboarding funnels become easy targets for automated compute arbitrage. Founders celebrating sudden spikes in signups risk footing massive GPU bills for zero-intent bots that will never convert.
- –Industrialized abuse: Farming operations have evolved beyond throwaway emails into automated tooling that spoofs machine GUIDs, clears local IDE state, and exploits temp-mail APIs.
- –Poisoned metrics: Heavy bot traffic artificially deflates activation rates and distorts cohort retention curves, masking genuine product-market fit signals.
- –Direct margin hits: Unlike traditional software with near-zero marginal cost per user, unvetted LLM queries directly burn through expensive frontier model tokens like Claude 3.5 Sonnet and GPT-4o.
- –Shifting trial design: AI startups are rapidly abandoning frictionless anonymous access in favor of upfront payment verification, reverse trials, and real-time device telemetry.
DISCOVERED
50m ago
2026-11-21
PUBLISHED
1h ago
2026-11-21
RELEVANCE
AUTHOR
finding_raj