Warp founder Zach Lloyd says an intrusion involving a vendor in Warp’s infrastructure reached a limited amount of agent trace and telemetry data. His October 26 statement says Warp detected the operation within hours on Friday and took steps to contain it.

Warp is a cloud-enabled developer terminal with AI features. Agent traces record activity from a model’s work with tools; telemetry records product activity. Lloyd’s statement does not identify the particular fields accessed or establish whether credentials, source code or command contents were included.

The company reports access; the attack explanation is preliminary

Evidence status: a vendor-reported intrusion and data access under investigation. Lloyd says Warp has notified everyone it has positively identified as affected. He describes “an upper bound of roughly 1% of Warp users” who may have been impacted.

That is the company’s provisional exposure estimate, not a confirmed victim count. The statement supplies neither its denominator nor an independently verified number of affected accounts.

Lloyd says the attack “appears to have come from an agent swarm”, likely beginning with a vendor dependency vulnerability. He provides no public technical evidence establishing that mechanism. Keep the reported data access separate from the hypothesis about how attackers operated.

The statement says Warp is working with CrowdStrike on a broader investigation and has engaged law enforcement. It names no affected supplier, CVE, customer configuration prerequisite or fixed client version. Its containment claim therefore cannot be translated into a customer patch matrix.

Follow account-specific instructions and clarify what the traces contained

For users and teams using Warp:

  • Review any incident notice through a trusted Warp channel. Lloyd says affected users will receive instructions describing the actions they need to take. The public statement contains no universal password-reset or key-rotation order.
  • Ask what data your account’s notice actually covers. Establish the affected time period, trace fields and integrations before deciding which downstream systems need investigation.
  • If the investigation identifies exposed credentials, revoke and replace those credentials. This is a conditional response to confirmed exposure, not a claim that Warp has announced secret theft.
  • Review future collection separately from incident recovery. Warp’s security page says users can opt out of telemetry while retaining its AI features. It also describes enterprise controls for telemetry, AI and secret redaction.

Those controls govern future use. Disabling telemetry now cannot withdraw data already accessed, and the reviewed sources do not show that opting out would have prevented this particular incident.

Caveat: The evidence is Warp’s preliminary statement, retrieved through a public post mirror because direct X extraction was blocked. The supplier, exact exposure and AI-attack attribution remain unresolved publicly.