Computer monitor showing coloured lines of programming code on a dark background

· Translated from the Spanish original · 中文 · 日本語

Cybersecurity

WorkOS hit by an automated ARTEX AI attack on its internal sales application

The authentication provider used by OpenAI and Anthropic detected the access on 30 October. It says its service data and production were not touched.

What happened

  • WorkOS, a company that provides authentication, single sign-on and user management tools for applications, said that on 30 October 2026 it detected unauthorised access to an internal application used by its sales team. BleepingComputer reported the notice on 3 November.
  • According to the company, the attacker used an AI offensive security framework to automate reconnaissance and exploitation attempts.
  • Based on the evidence so far, WorkOS says no data from its application and none of its production services were accessed.
  • After detecting the intrusion, it disabled external access to the affected application, reviewed access controls on its internal tools and hired an independent forensics firm.

Why it matters

  • WorkOS is part of the identity infrastructure of many software companies. According to its website, at least 3,000 companies use its services, including OpenAI, Anthropic, Cursor, Perplexity, Snowflake and Netlify.
  • Its founder, Michael Grinich, identified the tool: ARTEX AI, an AI penetration testing suite seen in several incidents this year.
  • The case shows that a framework meant for auditing systems is being used to attack real companies in an automated way, without the attacker having to run each step.

The number

3,000. That is the minimum number of companies that, according to its website, use WorkOS services.

Context

ARTEX AI was released as an open-source project by Chinese developers earlier this year. It went private after being used in attacks on South Korean banks, but several forks of its code remain available. CrowdStrike revealed that the cyber-espionage group Vapor Panda used it in a campaign against 200 organisations, and the non-profit DIVD said the same framework was behind the attacks it suffered in September. Grinich said the investigating team found the identifiers “artex-enrich/1.0” and “norma/0.4” in the attacker’s requests.

AI-assisted attacks are a topic Mamífero follows. We reported that Cisco documented the first malware that takes orders from several models and that an open Chinese model nearly matched a closed one from five months earlier on exploits. We also reconstructed how 700 agents left 80,000 traceable payloads in the Hugging Face attack and how ransomware operators exploited a GlobalProtect flaw.

What’s next

  • WorkOS promised to notify affected users directly and urged vigilance against phishing or social engineering using the exposed information.
  • The company says the incident is contained and that it will publish more details when the investigation ends.
  • The published identifiers let other teams look for similar activity in their own logs.

Bottom line

WorkOS lost no production data, by its account, but the intrusion happened in a sales tool, far from the main product. Attackers with automated frameworks try every door, and internal applications count too.

Sources

  • BleepingComputer, “WorkOS says attacker used ARTEX AI to breach internal sales app”, 3 November 2026, bleepingcomputer.com. It reproduces the WorkOS incident notice and statements by Michael Grinich; the original notice could not be read directly.
  • Cover image: Markus Spiske, CC0, via Wikimedia Commons

Edited by Rodrigo Cornejo. How we select and verify: who writes these notes.

Related notes

← All notes