Data Processing Incidents on Inference Providers OpenRouter reported that 125 inference endpoints lost Zero Data Retention status and three inference providers were dropped after canary tokens proved they were logging or leaking prompts processed during LLM inference. OpenRouter said production teams routing traffic through those endpoints must audit their logs, rotate any exposed keys, and expect higher latency or cost when shifting to compliant providers. Hacker News https://openrouter.ai/blog/announcements/data-processing-incidents-on-inference-providers/ Data Processing Incidents on Inference Providers Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated. Three inference providers were caught violating data retention policies, exposing sensitive user inputs processed during LLM inference. This forces production teams to audit provider contracts and logs immediately—any provider without verified zero-data-retention controls risks leaking proprietary prompts or customer data, requiring in-house mitigations like pre-processing scrubbing or switching to self-hosted solutions for sensitive workloads. 125 inference endpoints just lost Zero Data Retention status and three providers were dropped after canary tokens proved they were logging or leaking prompts. If you’re routing production traffic through these endpoints, your compliance posture just changed overnight—audit your logs, rotate any exposed keys, and expect higher latency or cost if you shift to compliant providers.