Nvidia told Brev customers that the SSH keys it sets up for their cloud machines may have been exposed during unauthorized activity from October 4 to 20; it has rotated them. Trigger.dev said on November 1 an attacker read environment variables for a subset of organizations. Both join an October run of attacks on AI developer platforms.
If your team runs GPU machines on Nvidia’s Brev or jobs on Trigger.dev, treat anything those platforms could reach as exposed and check it today. Nvidia says the SSH keys Brev sets up for customer machines may have been taken. Trigger.dev says an attacker read environment variables, the place most teams keep API keys and database passwords, for some of its customers. Both companies say they have shut the door. Neither has said how the attackers got in.
What changed
Nvidia Brev. Brev is Nvidia’s service for spinning up cloud GPU machines for AI development. In a notification to affected customers, quoted by Tom’s Hardware, Nvidia says information from its Brev network “may have been affected,” including the SSH key Brev configures for customer instances. Customers reported unauthorized activity from October 4 to October 20, a 16-day window. Per the notice:
- Nvidia rotated the affected SSH key and pushed the replacement to customer instances.
- It says it found no evidence the keys were misused.
- It has not said how the attackers got in, or whether they touched customer data or workloads.
Trigger.dev. The open-source background jobs platform posted a security disclosure on its status page on November 1. It says the exploit has been blocked, that the investigation found “exposure across a subset of organizations,” and that the attacker gained access to those organizations’ environment variables. Impacted account owners were emailed mitigation steps. The company declined further comment to The Register until it publishes a full report.
What to do now
For Brev:
- Look for the notice. If your account received it, your instances already carry a new key. Any copy of the old one on laptops, CI runners or shared vaults is dead weight; delete it.
- Audit the machines. Check
~/.ssh/authorized_keyson each instance for keys you did not add, and review login history for October 4 to 20. - Rotate what the machines held. Model API keys, cloud credentials, Hugging Face tokens and Git tokens stored on a Brev instance were reachable by anyone holding a working SSH key. Replace them.
For Trigger.dev:
- Read the email. Trigger.dev sent its steps only to affected account owners. If yours arrived, follow it first.
- Rotate every secret stored as an environment variable in your Trigger.dev projects, even if you got no email. The status page does not say how many organizations were affected.
- Check usage bills. Stolen AI API keys show up first as unexpected spend with your model provider.
Self-hosted Trigger.dev instances were not named in the disclosure; the incident page lists the hosted API service.
The October run against AI developer platforms keeps growing
Security firm Wiz’s “Black October” tracker lists Brev and Trigger.dev as the two latest publicly reported incidents. It counts nine other AI developer platforms hit in October, including Fireworks AI, Lightning AI, OpenCode, Railway, Warp and Northflank. Wiz says neither Brev nor Trigger.dev has shared attacker techniques or indicators yet, so there is no public evidence that one group is behind all of them.
The pattern so far is the same each time: the attackers go after the place where developers park credentials. Northflank told customers to assume every environment variable was decrypted, and stolen AI keys were used within days. If your team spreads secrets across several of these services, keep a list of where each key lives. It is the difference between a one-hour rotation and a week of guessing.