Chinese token resellers build discount LLM proxy market via fraud and credential pooling

The buried detail here is that the relay infrastructure itself is open-source and commoditized, meaning the barrier to standing up a new reseller node is essentially zero. Vendors are not fighting a few bad actors; they are fighting a replicable playbook anyone can fork.

This is largely disconnected from recent activity in our archive, as we have no prior coverage of API fraud, relay markets, or LLM reseller economics to anchor against. The story belongs to a broader conversation about inference pricing pressure that has been playing out across the industry: as frontier labs compete on cost-per-token, the spread between official pricing and marginal compute cost creates the exact arbitrage window this market exploits. The structural point, that tighter access controls and pricing recalibration are the only two levers available, is worth holding onto as a lens for evaluating any future vendor announcement about API tier changes or free-trial policy updates.

Watch whether OpenAI, Anthropic, or Google announce material changes to free-tier verification requirements or trial credit caps within the next two quarters. If those controls tighten in concert rather than unilaterally, it signals the providers have concluded the arbitrage problem is systemic and not solvable by one actor alone.