Good to knowAI agents

Two AI coding tools, Warp and OpenCode, reported breaches in October: agent logs and stored API keys were exposed. Rotate any secret an agent has seen

BleepingComputer reported on 27 Oct 2026 that Warp, an agentic terminal, lost agent decision logs and telemetry in an attack on a third-party vendor, with roughly 1% of users possibly affected. It says OpenCode told users the week before that stored API keys for outside AI providers were exposed, and some users saw unauthorized charges. What we know, what we do not, and what to check in your own setup.

How urgent

Good to know

Who is affected

Users of Warp (agent decision logs and telemetry, roughly 1% of users according to its CEO) and of OpenCode (stored API keys for outside AI providers, about 1.4% of users according to the report). Anyone who pastes keys or secrets into an AI agent session

What to do

If you use either tool, follow the vendor's email instructions and rotate any API key, token or connection string an agent could have seen, then check your provider's usage and billing for charges you do not recognize and set a spending limit. In general, keep secrets out of prompts and terminal output while an agent is running

Published 4 Nov 2026 · Sources last checked 4 Nov 2026No. 46 / 87

What happened

According to BleepingComputer (27 October 2026), Warp, an agentic development environment, disclosed that an attack on a third-party vendor in its infrastructure exposed agent decision logs (traces) and telemetry data. Warp's CEO said the company noticed the activity on a Friday and detected it within hours, and estimated that roughly 1% of users may be affected. Warp's website claims at least 800,000 users, which would be around 8,000 people, though the article does not state that figure. Warp says it notified affected users, brought in CrowdStrike and informed law enforcement.

The article adds that OpenCode, an AI coding platform, told users the week before that a breach exposed stored API keys for outside AI providers. About 1.4% of its users were affected and some saw unauthorized charges of up to $1,700.

What was exposed, and what was not

For Warp, the report says the company has not said whether source code, prompts or credentials were included. The vendor is not named and the weakness has not been described. So whether secretsSecret (API key, token, password) Any piece of text that proves to a service that you are allowed in, such as an API key, a database password or an access token. Tap again to open the glossary were exposed at Warp is an open question, while for OpenCode the exposure was API keys.

What to do

Our advice, based on the report:

  1. If you use either tool, read the email from the vendor and follow its instructions.
  2. Rotate secrets an agent could have seen: API keys, tokens and connection strings that you pasted into a prompt or that were printed in the terminal while an agent was reading it.
  3. Check usage and billing with your AI and cloud providers for charges you do not recognize, and set a spending limit and billing alerts.
  4. Be careful with unexpected "security" emails that claim to come from the vendor; check them against the vendor's own channels.
  5. Make it a habit: use keys with limited permissions, keep secrets out of prompts, and prefer environment variables or a secrets manager over pasting keys into a chat.

What we could not confirm

We read the BleepingComputer article. We did not read Warp's own statement (a post on X) or any OpenCode notice, so the numbers come from that one report. We do not know which vendor was attacked, how the attackers got in, what exactly was in the logs, or whether any Warp customer had secrets exposed. Estimated user counts are the vendors' own.

Comments

No account needed. Add a nickname if you like, or leave it empty to comment anonymously. Comments are reviewed before they appear. Links are not allowed.

    We store your comment, your nickname and a scrambled form of your connection that we use only to limit spam. We do not store your IP address or set cookies. Cloudflare Turnstile checks that you are not a bot.

    Sources and verification

    We use primary sources first and say when we rely on a news report. Dates are the day we last checked each source.

    Spotted a mistake? Email [email protected] and we will correct it and note the change.

    Get the free runbooks

    One-page security checklists for founders who build with AI. You get a confirmation email first, then the runbooks.

    Send me the runbooks