Warp, which builds a terminal with built-in AI agents, said attackers breached a third-party vendor in its infrastructure. From there they accessed AI agent execution traces and user telemetry data. CEO Zach Lloyd said the attack was detected within hours and that up to about 1% of users are estimated to have been affected. The company is working with CrowdStrike and has notified affected users and the authorities, but did not say whether code, prompts or credentials were exposed.

What it means: AI coding tools gather far more sensitive data than a simple chat history. An agent's execution traces can contain code snippets, terminal commands, environment variables, server names or even API keys that appeared on screen during the work. That is why a breach at a vendor hosting such data can have consequences well beyond the application itself. The incident also shows software supply chain risk: entry came through a third party, not Warp's own systems. For Greek development teams using Warp or similar tools, the practical step is to check whether they received a notice, rotate as a precaution any keys and tokens that may have appeared in agent sessions, and look in the settings at what telemetry is sent and whether it can be limited. Companies covered by the NIS2 directive, as transposed into Greek law, must also assess their suppliers, and AI developer tools now belong on that list, with clear limits on which data may leave the company.

Read the original on BleepingComputer

Related service: Custom software

Related stories