Breach · · 3 min read
Warp says a vendor breach exposed some AI agent traces and telemetry
Warp, the AI terminal for developers, says attackers got in through a vendor and reached agent trace and telemetry data for up to roughly 1% of its users.
Warp started as a terminal for developers and now lets people direct AI agents that write, test and run code. Its CEO says attackers got in through one of its vendors and reached some agent trace and telemetry data. If you use Warp, here’s what is known so far and a few sensible steps to take while the investigation continues.
What happened
In a statement posted on October 26, 2026, Warp CEO Zach Lloyd said that on Friday, October 23, the company became aware of “an active and narrow-banded exploitation of a vendor in Warp’s infrastructure.” He said Warp detected it within hours, fixed it and has been in touch with the vendor about the vulnerability, according to his post.
Lloyd said the attack appears to have come from “an agent swarm,” likely starting with a hole in a vendor dependency, and that it reached “a limited amount of agent trace and telemetry data.” Warp says it has notified everyone it has positively identified as affected, puts the upper bound at roughly 1% of its users, and is working with CrowdStrike and law enforcement, BleepingComputer reported.
Warp hasn’t named the vendor or the flaw. It also hasn’t said exactly what the exposed traces contained, so it’s not known whether they included code, prompts or any secrets that appeared in agent sessions.
What this means for you
Agent traces are records of what an AI agent was asked to do and what it did, including the commands it ran. In a developer tool, those records can show file paths, project names and sometimes keys or tokens that were pasted or printed along the way. That’s why it’s worth a quick check even if you haven’t heard from Warp.
What you can do
- Follow Warp’s instructions first if you got a notice. Lloyd says affected users will be told what action to take.
- Rotate any API keys, tokens or passwords that you pasted into, or that were printed in, a Warp agent session.
- Check activity on connected services such as GitHub, your cloud provider and AI model providers for anything you don’t recognize.
- Be careful with messages that claim to be from Warp. Go to Warp’s site or app yourself instead of following links.
- Keep secrets out of agent chats where you can. A secrets manager or environment file the agent can use without seeing keeps keys out of logs. A similar breach hit OpenCode users’ AI keys earlier this month.
How Frame helps
Add this breach to a verified email in Frame and Frame shows what it exposed, gives you a short checklist and counts it in your exposure report. Automatic breach lookups aren’t live yet. Frame’s email aliases also let you sign up for developer tools with their own address, so a leak at one doesn’t lead scammers straight to your main inbox. Frame never collects or stores the leaked records themselves.
Sources
- Zach Lloyd, statement on the Warp incident, October 26, 2026
- BleepingComputer, Warp discloses breach exposing AI agent logs, user telemetry data, October 27, 2026
New posts, by email.
About once a month. Plain-language privacy notes, nothing else. I never share your address.