Case · DiggingBeagle record

Botpress staff-session impersonation exposed customer AI bot configurations

An attacker exploited a staff SSO/proxy validation weakness to impersonate Botpress personnel and acquire a privileged employee API credential capable of reading or changing customer bot configurations. Botpress reports configuration reads across 16 workspaces and integration changes in two others during its October investigation.

Public disclosure: Nov 6, 2026

Scope

Reported incident or observed attacker activity; separate observed access from inferred downstream use and unverified victims.

Assessment profile

Assessment method
Setting
production
Exploitation
confirmed in the wild
Evidence
Grade B
Remediation
fixed reported
AI role
AGAINST AI
Basis and provenance

production; confirmed in the wild. The cited primary disclosure supports the specific sequence and boundaries described here: Attacker exploits weak staff sign-in proxy validation and explores Botpress internal tools beginning October 12. The evidence does not independently establish consequences beyond the record's supported outcome; a numerical impact band cannot be justified from the available evidence. Evidence grade B reflects the source attribution and independent verification limits.

Assessed Nov 7, 2026 using diggingbeagle.assessment/1.

At a glance

Mechanism and trust boundary

  1. 01

    Attacker exploits weak staff sign-in proxy validation and explores Botpress internal tools beginning October 12.

  2. 02

    Internal access yields a staff API credential already authorized in some customer workspaces.

  3. 03

    On October 21, the credential reads configurations in 16 workspaces and modifies integrations in two additional workspaces.

  4. 04

    Botpress patches the proxy and revokes the credential, but customer-side integration secrets may still need separate rotation.

Timeline

  1. Oct 12, 2026
    Event type unspecified

    Authentication proxy probes begin

    Botpress reports anomalous staff authentication and proxy-validation activity preceding the principal customer configuration access.

  2. Oct 21, 2026
    Event type unspecified

    Customer configuration access

    Between 07:44 and 08:23, attacker access reached 16 customer workspace configurations; integrations in two additional workspaces were modified.

  3. Nov 3, 2026
    Event type unspecified

    Attempted re-entry blocked

    Botpress reports later attempted re-entry through November 3 after mitigation, without treating each attempt as successful access.

  4. Nov 6, 2026
    disclosure

    Public disclosure

Claims & evidence

CLM-DB07-009-OUTCOMEBotpress reports configuration reads across 16 workspaces and integration changes in two others during its October investigation. Potentially reachable AI or CRM secrets are not equivalent to confirmed theft of every token or user conversation.supported

Basis: reported finding

Link to claim
CLM-DB07-009-MECHANISMAn attacker exploited a staff SSO/proxy validation weakness to impersonate Botpress personnel and acquire a privileged employee API credential capable of reading or changing customer bot configurations.supported

Basis: reported finding

Link to claim
CLM-ENR-DB07-012-BOTPRESS-CONFIG-TIMINGBotpress reports that the attacker accessed configurations across 16 customer workspaces and altered integrations in two other workspaces during the October 21, 2026 intrusion window from 07:44 to 08:23. That observation does not establish theft of every reachable integration secret.supported

Basis: reported finding

Link to claim

Implications

Check token use and revocation at each downstream AI provider and CRM, not only staff-session termination.

Unknowns and contradictions

  • 16 read+2 modified not all confirmed secret theft. No published overall money lost, compromised user conversations or verified resale.

Sources and citation

Material revision history

  1. Nov 7, 2026 · Canonical change recorded · new in release · revision 227