---
title: 'In the News: November 21, 2026 (Extra 3)'
description: 'Together AI says an AI coding agent committed the credential behind its prompt leak; OpenRouter has cut three providers and 125 ZDR endpoints.'
canonical_url: 'https://darkfactory.dev/news/2026-11-21-extra-3'
markdown_url: 'https://darkfactory.dev/news/2026-11-21-extra-3.md'
collection: news
date_published: '2026-11-21T23:01:00-05:00'
date_modified: '2026-11-21T23:01:00-05:00'
---

# In the News: November 21, 2026 (Extra 3)


The providers named in this week's canary-credential research have started publishing their own accounts. The most useful one for anyone running coding agents comes from Together AI: the credential that exposed its stored prompts was committed by an AI coding agent. OpenRouter, which routed much of the affected traffic, has also said what it removed.

## 1. Together AI says an AI coding agent committed the credential behind its prompt leak

**[Security event: Prompt leakage for requests in June and July of 2026](https://www.together.ai/blog/security-event-prompt-leakage-for-requests-in-june-and-july-of-2026)** · Vipul Ved Prakash, CEO and co-founder, Together AI · Together AI blog, November 21, 2026

Together stored prompt samples for software testing in a private Hugging Face repository. The samples came from requests to MiniMax M3, GLM-5.2 and Kimi K2.6, mainly on June 24, July 4, July 8 and July 14, 2026. The post says that while the company was developing its public GitHub repository, xorl, "an AI coding agent committed a credential that provided access to the private Hugging Face repository." Automated scanning caught the credential and the commit was reverted, but during the exposure window an unauthorized party took it, downloaded the samples, and searched them for credentials to other services. Five of the credentials it tested were canaries from [the Dragos experiment](https://www.dragos.com/blog/ai-supply-chain-security-canary-tokens).

Together lists four changes: scoped or short-lived credentials and keyless access where supported, credential checks on every pull request, zero data retention (ZDR) as the default for new accounts, and updated development processes for AI coding agents. The post does not say which agent was involved or what the new agent processes are. Its review of what the prompts contained is still in progress.

**Why it matters:** This is a provider's own post-mortem in which the leaked credential was committed by an agent. Scanning caught the secret after it landed, and that was not fast enough. If your agents can commit to a public repository, put the secret check in front of the commit, and keep the credentials those agents can read short-lived and narrowly scoped.

## 2. OpenRouter removes three providers and 125 endpoints from zero data retention

**[Data Processing Incidents on Inference Providers](https://openrouter.ai/blog/announcements/data-processing-incidents-on-inference-providers)** · OpenRouter · OpenRouter blog, November 22, 2026

OpenRouter says it has off-boarded three providers and removed 125 endpoints from ZDR eligibility. The removals cover the providers researchers flagged plus further endpoints taken out as a precaution, and all potentially affected customers have been notified. The post does not name the three off-boarded providers. OpenRouter says it is adapting the researchers' tests to run continuously, checking network paths for requests forwarded to third parties, and will require "appropriate assurances and validation" before any removed endpoint gets ZDR back. It also says the findings showed no breach of OpenRouter itself.

It links three provider notices. [Novita](https://novita.ai/blog/minimax-requests-privacy-notice/) says a routing configuration issue sent some MiniMax requests to MiniMax's official API, so requests between 2026-05-28 and 2026-11-11, including those routed through OpenRouter, "may have been accessed or retained by a third party." [Phala](https://phala.com/posts/phala-november-preliminary-incident-report), in a preliminary report dated November 10, tells users whose DeepSeek V4 Flash 0731 requests OpenRouter routed to it between October 21 and November 8 to treat those requests as exposed and rotate any secrets they contained. The Hacker News thread had 42 points and 12 comments at 04:02 UTC on November 22.

**Why it matters:** A ZDR label on a routed endpoint is a policy claim, and OpenRouter has now pulled it from 125 endpoints until they can be validated. If your agents send repository contents through a router, check which endpoints your traffic used in the exposure windows above, and rotate anything sensitive that was in those prompts.
