AI-Linked Breaches Mount as WorkOS, Northflank, Trigger.dev Disclose Hacks

What it is

SecurityWeek reports that at least a dozen AI developer platforms were breached in about a month (to early November 2026), with attackers mainly seeking stored customer secrets, in particular AI provider access keys. Named victims beyond the summary's list: Andromeda, Brev (NVIDIA), Fireworks AI, Lightning AI, OpenCode, Railway, Warp and Zeabur. The attacks are described as part of one trend but not as one campaign. Initial access varies: compromised credentials, third-party dependencies and vulnerability exploitation.

How it works

The Cloud Security Alliance says intruders reach AI provider credentials through intermediary layers: hosting platforms, evaluation services and developer consoles. These hold keys on behalf of many downstream customers, so one intrusion can expose keys for many inference accounts at once.

Wiz reports abuse of enterprise credentials stolen in these breaches. Two cases showed the same first unauthorized activity within 30 minutes of the reported credential access inside the vendor breach. Several threat actors used the same compromised keys, which Wiz says suggests a robust market. Wiz also cites distillation demand, "notably alleged against Chinese AI labs".

Preconditions / Affected

You are exposed if you stored AI provider keys or other secrets with any of the named platforms, or with a similar host, gateway or console.

• Chainstack (attacked 26 October): access came through a third-party component into a development environment and parts of production. The attacker reached a database with organization names, account holder names, member email addresses and password hashes. Node credentials may also be exposed.
• WorkOS (attacked 30 October): an internal sales application was compromised. Exposed data includes contact details, job titles, customer account histories, summaries of customer communications and information on customers' use of WorkOS products.

What to do

The page gives one explicit action: Chainstack asked customers to rotate node credentials. The page states nothing further. Rotating AI provider keys held by any named platform follows from the 30-minute abuse window Wiz observed, but that is an inference, not a page recommendation.

Caveats

• The page does not say which platforms confirmed key theft versus other data loss, or how many keys were taken.
• Artex is an open source AI pentest framework. The page says multiple companies linked Storm-0940 attacks (attributed to a Chinese APT) to Artex use, including a ConnectWise customer campaign. It does not say Storm-0940 is behind Chainstack or WorkOS, only that those two attacks were automated with Artex.
• The Wiz forecast that more actors will turn to credential theft is a projection, not an observation.