OpenRouter drops 3 providers, pulls ZDR from 125 endpoints
OpenRouter off-boarded three inference providers and removed 125 endpoints from zero data retention after Dragos canary research. What to check in your stack.
The router took the hit for its providers. On November 22, OpenRouter announced that it off-boarded three inference providers and removed 125 endpoints from zero data retention (ZDR) eligibility. The trigger was the Dragos canary study we covered yesterday. If your app sends prompts through OpenRouter, the provider pool behind your API key changed this weekend.
What actually happened
Per OpenRouter's post:
- Three providers are gone. OpenRouter off-boarded them based on its own investigation. It does not name them.
- 125 endpoints lost ZDR eligibility. That count includes endpoints flagged by the researchers plus others pulled "as a precautionary measure." Do not read 125 as 125 confirmed leaks.
- OpenRouter itself was not breached. The company says the findings showed no breach or unauthorized access of OpenRouter.
- Affected customers were notified. If you did not get an email, OpenRouter does not consider you affected.
- ZDR comes back only with proof. OpenRouter says it needs "appropriate assurances and validation" before it restores any removed endpoint. No timeline.
- New checks. OpenRouter says it will run the researchers' canary-style tests on an ongoing basis and verify network paths to catch providers that forward requests to third parties.
The post links incident notices from three providers: Novita, Phala, and Together AI. It does not say which providers it off-boarded, and it says not every affected provider has posted publicly.
Why it matters for your business
Routers sell one thing: you write one integration and get every model at the best price. The cost is that you stop knowing who runs your request. A ZDR checkbox told you the provider promised not to keep your data. Dragos showed some promises did not hold. OpenRouter's response is the right one, but it is reactive. It found out from a third party.
Two practical effects. First, if you require ZDR, your routable pool just got smaller. Expect different latency, different prices, or failed requests on niche open-weight models with few hosts. Second, if your prompts carry anything sensitive, the question is no longer "is ZDR on?" It is "which company ran my request last Tuesday, and how would I know?"
What we would do this week:
- Pull your activity logs and list which providers served your traffic from July to November.
- Pin providers for sensitive workloads instead of letting the router pick the cheapest one.
- Retest your fallbacks now that the pool has changed.
- Keep credentials out of prompts. No router setting fixes a key that is already in the context window.
Key takeaways
- OpenRouter off-boarded three unnamed inference providers on November 22, 2026
- 125 endpoints lost zero data retention eligibility, some only as a precaution
- OpenRouter says it was not breached and has notified affected customers
- Removed endpoints get ZDR back only after "assurances and validation," with no timeline
- If you require ZDR, retest routing and fallbacks; the provider pool changed
Do you know which company ran your last prompt? We audit AI routing, pin sensitive workloads to providers you have a contract with, and move secrets behind a gateway so a provider breach is not your breach. See what we build or book an AI stack review.
Sources: OpenRouter: Data Processing Incidents on Inference Providers, Dragos: Your AI supply chain needs canaries.
- #openrouter
- #zero-data-retention
- #ai-security
- #ai-routers
- #prompt-privacy
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
OpenAI Performance Factory: Codex agents debug, engineers merge
OpenAI's Performance Factory ran 1,700+ Codex investigations into ChatGPT slowdowns. 70% of audited root causes held up; about 15 agent PRs shipped as written.
Read itGemini API Tool Retrieval: stop paying for tool schemas
Gemini API Tool Retrieval is in preview: agents load tool schemas on demand with defer_loading, at no extra charge. Here's when it saves tokens and when it won't.
Read it