{"slug": "openrouter-warns-minimax-prompts-may-have-been-read-zdr-labels-removed", "title": "OpenRouter Warns MiniMax Prompts May Have Been Read: ZDR Labels Removed", "summary": "OpenRouter emailed paying customers on November 21, 2026 warning that prompts routed to MiniMax models between July and October 2026 may have been read, after a security researcher's test URLs embedded in those prompts were fetched by a third party 40 to 130 hours after the requests completed. OpenRouter stripped Zero Data Retention (ZDR) badges from the affected MiniMax endpoints while it investigates, and named AtlasCloud, Novita and StreamLake as hosts that forwarded MiniMax traffic to the Chinese lab; OpenRouter said it had not received MiniMax's own account of events as of the email. The notice follows a November 19 Dragos study by HD Moore that seeded roughly two million prompts with canary tokens to trace where secrets resurfaced.", "body_md": "**OpenRouter began emailing paying customers on November 21, 2026 with a blunt subject line: requests you sent to MiniMax models may have been read.** The message ties a third-party access pattern to prompts routed to MiniMax between July and October 2026, names resellers that forwarded traffic to the Chinese lab, and says OpenRouter stripped **Zero Data Retention (ZDR)** badges from the affected MiniMax endpoints while it investigates. There is still no public blog post on [openrouter.ai/blog](https://www.openrouter.ai/blog?ref=explainx); the notice lives in inboxes first.\n\nThis is the customer-facing half of a story [Dragos published two days earlier](https://explainx.ai/blog/dragos-hd-moore-two-million-canary-prompts-three-provider-breaches-zdr-ignored-ai-routers-2026): HD Moore seeded roughly two million prompts with **canary tokens** and watched where secrets resurfaced. Builders who routed agents through OpenRouter for cheap MiniMax inference now need a rotation plan, not a headline recap.\n\n## TL;DR\n\n| Question | Answer (Nov 21, 2026) | \n|---|---|\n| What happened? | Customer email: test URLs in MiniMax prompts were opened **40–130 hours** after inference | \n| Who might have read prompts? | OpenRouter blames MiniMax and hosts that forwarded to it, not its own core platform | \n| What changed on OpenRouter? | **ZDR labels removed** from MiniMax-model endpoints pending investigation | \n| Resellers named | **AtlasCloud, Novita, StreamLake** said they forward MiniMax traffic to MiniMax | \n| Research backdrop | [Dragos canary study](https://explainx.ai/blog/dragos-hd-moore-two-million-canary-prompts-three-provider-breaches-zdr-ignored-ai-routers-2026) (Nov 19) on prompt theft and ignored privacy flags | \n| What to do | Rotate keys, audit July–Oct MiniMax usage, avoid sensitive prompts on marketplace routes | \n\n## What OpenRouter told customers\n\nSeveral developers posted the full text of OpenRouter's email on [Hacker News on November 21, 2026](https://news.ycombinator.com/item?id=50585962&ref=explainx). The message is personalized: it lists how many requests **your** account sent to MiniMax models, which upstream hosts carried them, and the UTC date range.\n\nThe facts OpenRouter attributes to the unnamed security researcher are specific:\n\n- Unique **test URLs** were embedded in prompts aimed at**MiniMax models** from**July through October 2026** .\n- A **third party** fetched those URLs**40 to 130 hours** after the requests completed.\n- The pattern held whether **MiniMax answered directly** or another provider**forwarded** the call to MiniMax.\n- Parallel tests on **other models** at the same hosts were**mostly untouched** , and**none** of the planted AWS keys were used (per the email text reporters shared).\n\nOpenRouter's conclusion in the email is explicit: based on provider reports, the problem sits with **MiniMax** and the companies that **forwarded** OpenRouter traffic to it, **not** with OpenRouter's own routing core. That wording matters for legal and compliance teams, but it does **not** erase the prompts that already left your environment.\n\nWhile investigating, OpenRouter says it **removed the Zero Data Retention label** from the MiniMax-model endpoints in question. That is a product change paying users can see in the model catalog today even without the email.\n\n### Providers in the forwarding chain\n\nThe email names three hosts that told OpenRouter they forward MiniMax listings to MiniMax itself:\n\n1. **AtlasCloud**\n2. **Novita**\n3. **StreamLake**\n\nIt also relays MiniMax's statement to StreamLake: **content moderation** was enabled on MiniMax endpoints **before October 22**, and MiniMax has since enabled **Zero Data Retention** on its side. OpenRouter says it asked MiniMax for its **own account** of events and had **not received one** as of the email send.\n\nNone of those sentences appear on a public OpenRouter status page yet. For how OpenRouter describes retention tiers when they **are** honored, see its [data collection documentation](https://openrouter.ai/docs/guides/privacy/data-collection?ref=explainx).\n\n## Why the delay matters more than the brand name\n\nDragos did not name the Chinese lab in its [November 19 post](https://www.dragos.com/blog/ai-supply-chain-security-canary-tokens?ref=explainx). Reporters covering the OpenRouter email note the same **July–October** window and **multi-day** lag between prompt and URL fetch. A real-time content filter might hit a link within seconds. **Days** fit **stored prompts** being opened later.\n\nMoore wrote that prompts sent to **U.S. hosting providers reselling the lab's models** still triggered the **same scraper**, meaning traffic marketed as domestic could **flow back** to the lab. OpenRouter's email about **forwarded MiniMax routes** rhymes with that paragraph even though neither document cross-links.\n\nIf you already read our Dragos summary, treat this email as **action items for OpenRouter accounts**, not a second research paper.\n\n## Zero Data Retention versus Safety Retention\n\nVault door with a key hanging beside it, suggesting stored secrets at risk\n\nOpenRouter's docs draw a line builders often miss when they tick a privacy box:\n\n| Setting | What OpenRouter says it means | \n|---|---|\n| **Zero Data Retention (ZDR)** | Prompts and completions should not be stored after processing, aside from required safety paths | \n| **Safety Retention** | Provider may retain prompts to prevent abuse even when you asked for no training use | \n\nThe MiniMax email concerns **ZDR labels** on specific endpoints. Removing the badge does not prove every host was malicious; it signals OpenRouter will not promise ZDR on those routes **until** the investigation closes.\n\nMoore's blunt summary still applies: **a privacy flag only binds the first hop.** Our [semantic caching and router cost guide](https://explainx.ai/blog/semantic-caching-llm-apps-cost-latency-guide-2026) explains why multi-hop inference is cheap and why cheapness can hide extra readers.\n\n## How this fits the wider AI supply chain story\n\nThe MiniMax email is one tile in a busier November:\n\n- **Dragos (Nov 19):** three provider**breaches** with stolen canary AWS keys,**marketplace resellers** logging prompts, and**ignored ZDR** paths, from ~two million seeded prompts.\n- **[Railway's October secret leak](https://explainx.ai/blog/railway-openrouter-attacks-ai-api-keys-stolen-what-to-do-2026):** attackers hunted**AI API keys** in hosting env vars; OpenRouter spend was in the reimbursement table.\n- **[Google GTIG on LLMjacking](https://explainx.ai/blog/google-gtig-llmjacking-llm-mixers-1-million-stolen-cloud-ai-access-what-to-do-2026):** stolen cloud AI keys turned into**mixer** services reselling access.\n\nOpenRouter is also the router behind [Hugging Face Inference Providers since November 16](https://explainx.ai/blog/hugging-face-openrouter-inference-routing-november-2026). That partnership does not cause this incident, but it shows how central the router is for everyday model access.\n\nFor background on why developers flocked to MiniMax on routers, see our [MiniMax M3.1 Flash pricing post](https://explainx.ai/blog/minimax-m3-1-flash-weights-oct-23-space-bunny-benchmarks-pricing-2026) and the [OpenRouter default routing change](https://explainx.ai/blog/openrouter-default-routing-quality-over-price-fewer-fallbacks-429s-open-weight-models-claimed-vs-verified-2026) from October.\n\n## What developers should do this week\n\n### 1. Read your email, then read your logs\n\nOpenRouter's message includes **per-account** request counts and hosts. Match that window against:\n\n- CI jobs that sent **repo context** through MiniMax\n- Agent harnesses with **long system prompts** holding credentials\n- Experiments that assumed **ZDR** because the catalog said so\n\n### 2. Rotate secrets that lived in prompts\n\nGreen key enclosed in a capsule beside a closed folder\n\nIf a prompt could contain **API keys**, **database URLs**, **customer PII**, or **unreleased code**, rotate those assets even when the email says AWS canaries were untouched. Canary keys prove **read access**, not the only thing worth stealing.\n\n### 3. Pick routes deliberately\n\nMoore's checklist from Dragos still fits:\n\n- **Self-host** or go**direct** to a first-tier provider for sensitive work.\n- On routers, **choose upstream providers** instead of accepting the cheapest pool.\n- **Seed your own canary tokens** in agent context and treat firings as incidents.\n\nFor agent pipelines that auto-call tools, add guardrails before outbound network calls. [AgentBeam](https://agentbeam.com/?ref=explainx) can **stop agents from exfiltrating secrets** once a model or router misbehaves.\n\n### 4. Watch for a public post-mortem\n\nAs of this publish time, explainx.ai sees **email plus docs**, not a blog post. Check [openrouter.ai/blog](https://www.openrouter.ai/blog?ref=explainx), your OpenRouter dashboard notices, and whether **ZDR badges return** on MiniMax listings.\n\n## Is OpenRouter still worth using?\n\nFor many builders, yes, with eyes open. OpenRouter's value is **one key, many models**, as we covered in the [enterprise OpenRouter guide](https://explainx.ai/blog/what-is-openrouter-enterprise-guide-2026). This incident does not delete that convenience. It does mean **model badges are claims you must verify**, especially on **discounted marketplace** listings that may forward overseas.\n\nTreat the November 21 email like a **billing anomaly alert**: investigate, rotate, document, then decide whether MiniMax routes belong in production at all.\n\n## What we do not know yet\n\nGaps worth stating plainly:\n\n- OpenRouter has **not** published the researcher name; reporters match the details to**Dragos** , but neither party has confirmed that link on the record.\n- **MiniMax** has not issued a statement OpenRouter could quote directly, only a relay through StreamLake about moderation and ZDR.\n- Some outlets report **longer windows** (for example into November) for specific accounts; reconcile against**your** email, not forum anecdotes.\n- Removing ZDR labels is **not** the same as proving every prompt was read; it is OpenRouter pausing a promise until evidence is sorted.\n\n## Is OpenRouter down or broken?\n\nSome searchers confuse product incidents with availability. As of November 21, 2026, explainx.ai saw **no OpenRouter status-page outage** tied to this email. The warning is about **historical MiniMax routes** and **ZDR label changes**, not a platform-wide failure. If API calls fail today, check [OpenRouter status](https://status.openrouter.ai/?ref=explainx) separately from this privacy story.\n\n## Questions paying teams should ask procurement\n\nSecurity review templates rarely mention **inference marketplaces**. After the MiniMax email, ask vendors:\n\n1. **Which upstream host** runs each model SKU in production, not only the router brand.\n2. Whether **ZDR** is contractual or a catalog badge that can disappear overnight.\n3. Whether **reseller forwards** to non-US entities are disclosed in your DPA.\n4. How **agent frameworks** log prompts when they fan out to five models for one task.\n\nThose questions pair with our [token fraud and free-trial abuse](https://explainx.ai/blog/token-fraud-chatprd-claire-vo-free-trial-abuse-stripe-radar-ai-startups-2026) coverage: cheap inference often funds itself with **data you did not price in**.\n\n## MiniMax-specific background for new readers\n\n[MiniMax](https://explainx.ai/blog/minimax-m3-1-flash-weights-oct-23-space-bunny-benchmarks-pricing-2026) is a Chinese lab whose models became router favorites when **Space Bunny** topped usage charts. Low per-token pricing makes MiniMax attractive for **batch translation, agent loops, and coding harnesses** that burn millions of tokens. That economics story does not change because of one email, but it explains why so many accounts touched MiniMax in July through October even if they never chose the lab directly.\n\n## A simple rotation checklist (30 minutes)\n\n| Step | Action | \n|---|---|\n| 1 | Export OpenRouter usage or read the email counts for **MiniMax** models | \n| 2 | List secrets that appeared in those prompts (keys, URLs, customer rows) | \n| 3 | Rotate each secret at the **source system** , not only on OpenRouter | \n| 4 | Create a **new OpenRouter key** with spend caps for remaining tests | \n| 5 | Switch sensitive workloads to **direct Anthropic or OpenAI** routes temporarily | \n| 6 | File a ticket if your compliance team needs OpenRouter's written confirmation | \n\nIf you route through [Bedrock tokenomics dashboards](https://explainx.ai/blog/amazon-bedrock-tokenomics-data-lake-finops-2026), add a FinOps note so finance sees why November spend shifts providers.", "url": "https://wpnews.pro/news/openrouter-warns-minimax-prompts-may-have-been-read-zdr-labels-removed", "canonical_source": "https://explainx.ai/blog/openrouter-minimax-prompts-may-have-been-read-customer-email-november-2026?utm_source=rss&utm_medium=feed&utm_campaign=blog", "published_at": "2026-11-21 00:00:00+00:00", "updated_at": "2026-11-21 17:42:29.585596+00:00", "lang": "en", "topics": ["ai-safety", "ai-policy", "ai-infrastructure", "large-language-models"], "entities": ["OpenRouter", "MiniMax", "AtlasCloud", "Novita", "StreamLake", "HD Moore", "Dragos"], "also_reported_by": [], "takeaway": null, "recommendation": null, "alternates": {"html": "https://wpnews.pro/news/openrouter-warns-minimax-prompts-may-have-been-read-zdr-labels-removed", "markdown": "https://wpnews.pro/news/openrouter-warns-minimax-prompts-may-have-been-read-zdr-labels-removed.md", "text": "https://wpnews.pro/news/openrouter-warns-minimax-prompts-may-have-been-read-zdr-labels-removed.txt", "jsonld": "https://wpnews.pro/news/openrouter-warns-minimax-prompts-may-have-been-read-zdr-labels-removed.jsonld"}}