WorkOS has disclosed unauthorized access to an internal sales application that exposed customer contact details and records of business conversations. The company detected the security incident on October 30, 2026, and released a security bulletin on November 2 after identifying an AI-powered offensive security framework used to automate the attack.
Automated reconnaissance and exploitation framework
The incident involved the use of an AI-assisted framework that automates reconnaissance and exploitation attempts. WorkOS founder and CEO Michael Grinich identified the tool as ARTEX, a multi-agent platform that automates offensive security tasks. The platform turns broad attacker goals into specific commands, allowing its agents to test permissions, map networks, and adjust failed commands across several attack stages with limited human intervention.
Information exposed in the breach
The affected application contained several categories of information related to customer relationships and account histories. Data points available through the application included customer names, email addresses, phone numbers, job titles, and company affiliations. The system also held sales relationship information and summaries of customer account histories.
Customer communications and operational metadata
Records within the application included excerpts and summaries from sales calls, support discussions, and conversations held in customer Slack channels. Beyond communication records, the breach exposed operational metadata regarding how customers use WorkOS products, including connection health, directory status, and general product usage statistics. WorkOS noted that these categories represent the types of information available in the application rather than a confirmed record of every specific item accessed.
Scope of the investigation
Current investigative findings indicate that the unauthorized access was limited to CRM and customer support information. There is no evidence that WorkOS production services or application data were accessed or impacted. The company is currently reviewing affected records to determine the specific impact on individual organizations and has retained an independent digital forensics and incident response firm to assist the investigation.
Context of automated security tool abuse
The incident reflects a growing trend in the abuse of multi-agent platforms for automated security testing. Previous research by Wiz and Google Threat Intelligence Group has documented ARTEX agents performing Active Directory discovery, credential attacks, and lateral movement. Observations from other investigations have shown the framework downloading application files, discovering internal API routes, and testing authentication controls. While these capabilities have been documented in other cases, they are not confirmed steps in the WorkOS breach.
Security response and customer advice
Following the discovery of the incident, WorkOS disabled external access to the compromised sales application and began a review of access controls across all internal tools. The company stated it will contact affected customers directly once findings are verified. Because exposed conversation details can be used to create more convincing phishing messages, WorkOS advises customers to verify any unexpected requests before sharing sensitive data, even if the messages reference real colleagues or legitimate support discussions.
Licensed under CC BY-ND 4.0 International