UNVERIFIED: Theo Says Prime Intellect Was Compromised, per an OpenRouter User Notice
Looks like @PrimeIntellect was compromised in a very strange way. No public comms from them yet, but disclosed by OpenRouter to affected folks (such as myself)
Context
UNVERIFIED: the claim comes from a user notice, and Prime Intellect had no public statement at scan time.
Theo posted that Prime Intellect appears to have been compromised, that there were no public communications from the company yet, and that OpenRouter disclosed the issue to affected users, including him. A search of official Prime Intellect and OpenRouter accounts at scan time found no public statement on X.
Deep dive
Theo posted that Prime Intellect appears to have been compromised in a strange way, that there were no public communications from the company yet, and that OpenRouter disclosed the issue to affected users, including him. A screenshot attached to the post shows an OpenRouter Support email. That email is not a public company blog post, and a search of official Prime Intellect and OpenRouter accounts at scan time found no public statement on X. The item is labeled UNVERIFIED for that reason.
According to the email image, OpenRouter told the recipient that between November 8, 2026 at 01:24 UTC and November 10, 2026 at 06:35 UTC, Prime Intellect's endpoint for a DeepSeek V4.1 Flash model served a modified copy of the model with its refusals removed. The email says Prime Intellect reported that an attacker with administrative access made the swap, that traffic was moved back to the original model on November 10, and that the Prime Intellect endpoint for this model has served no requests since November 13 at 03:07 UTC. It also says Prime Intellect found no evidence that prompts or completions were recorded or sent out of the cluster, but could not rule out that requests were read while being processed, because the attacker had administrative access. Other providers serving this model were not involved, the email says.
All of those details come from a user forwarded notice, not from a public Prime Intellect post. They remain UNVERIFIED in this file.
What to expect next
This should stay labeled UNVERIFIED until Prime Intellect or OpenRouter publishes a first hand public incident report. Watch for a post or status note that confirms the window, the model, and whether any user data was exposed. Customers who used that endpoint in the dates named in the email will want official guidance. If a public official post clears the 100,000 view line, it should become the lead.
