Observed incident

Botpress reports unauthorized configuration access in 16 customer workspaces

Reported

Published here

What happened

An intruder obtained a staff API credential and used it on October 21 to read bot configurations in 16 customer workspaces. The intruder also changed one integration in each of two other workspaces. Configuration settings can contain AI provider keys and other service credentials.

Parties and reporting

AI tools and systems: Botpress AI agent platform, Workspace integrations and staff API credentials

Outcome

Botpress fixed the sign-in flaw, revoked the staff credential and notified affected owners. It says subsequent attempts it observed through November 3 failed. Internal credential rotation and hardening continued at publication.

Operating lesson

Rotate each credential stored in an affected bot at its issuing provider. Review staff membership in customer workspaces and check provider logs for unauthorized use.

Evidence and limits

Botpress assesses that the attacker sought AI usage for resale. Its logs do not show which bots held valid credentials. Botpress found no evidence of access to end-user conversations or publication of customer data.

Sources