Key Info

OpenCode is moving toward running 100% of its inference on its own GPUs, with dax citing a study that seeded 2 million canaries into popular AI services and tracked their theft and abuse.

Highlights

  • The study identified three compromised providers, resellers stealing prompts via built-in logging, and providers ignoring ZDR and privacy flags.
  • dax argues gateway products talk about enterprise security and privacy but do not actually control anything.
  • The move to self-hosted GPUs is framed as a direct response to those privacy and control gaps.