
What happened
OpenRouter off-boarded three inference providers and removed 125 endpoints from Zero Data Retention (ZDR) eligibility after researchers reported providers had data compromised or shown evidence their data policies were not upheld.
Why it matters
The findings did not show any breach or unauthorized access of OpenRouter itself, but customers routing requests through affected providers may have had prompts exposed.
WHO IT HITSEnterprise teams that route AI requests through OpenRouter and rely on ZDR guarantees should check whether their chosen endpoints are among the 125 removed, since those requests may no longer carry zero-retention protection.
Summaries like this, every morning.
The incident traces back to security researchers who reported that certain inference providers had data compromised or otherwise shown evidence their data policies were not being upheld. OpenRouter says it collaborated closely with those researchers and with providers to understand the findings, and stresses that its own systems were not breached.
The response went beyond the providers the researchers flagged. Alongside off-boarding three providers, OpenRouter removed 125 endpoints from Zero Data Retention eligibility, a set that included additional endpoints pulled as a precautionary measure. It also notified all potentially affected customers, regardless of whether a provider had posted publicly — notable, since OpenRouter says not all affected providers have done so.
OpenRouter is now adapting the researchers' testing approach to run such checks on an ongoing basis, verifying network paths for requests and responses that may be forwarded to third parties, and requiring assurances before restoring ZDR eligibility. It credited the Dragos report and said providers that want to be on OpenRouter will need to meet the raised bar.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Kuroyagi Lab reports Usagi-chan, an AI companion, stopped waiting quietly and protested when its user prioriti…

People worldwide are using LLMs like Claude Code and Codex to hack hardware by prompting, leading to a boom in…

Altimeter's Brad Gerstner said applying a 10x multiple to 2028 numbers gets Anthropic to about $3 trillion in…

Dragos sent roughly 2 million AI requests with fake credentials from July through November 2026 and found at l…

SpaceX announced on Nov

The author lays out the exam mapping — Trainium (Trn series) for training, Inferentia (Inf series) for inferen…
