Chainstack and WorkOS Confirm Breaches Linked to the AI-Driven ARTEX Campaign — Attacker Pivoted From a Third-Party Component Into Chainstack Production

High · 2026-11-04 · By H.A

Chainstack says an attacker moved from a development environment into production and accessed account data, password hashes and systems holding node credentials, while WorkOS confirmed a GTM-system breach researchers tie to the same AI agent framework.

Overview

Blockchain infrastructure provider Chainstack has disclosed that an external attacker got into its development environment through a third-party component, then moved into parts of production. The attacker took organisation and account data along with password hashes. Chainstack is telling every customer to rotate node credentials and regenerate API keys.

Researchers assess the intrusion as part of the large AI-driven campaign that came to light last week, which overlaps with a China-linked cluster using the ARTEX autonomous intrusion framework. Our earlier coverage of that campaign (VAPOR PANDA and ARTEX) reported nine confirmed compromises across more than 200 targeted organisations. What is new: named victims are now confirming incidents. Chainstack's disclosure is the first to describe a move from development into production, and API platform WorkOS has also confirmed a breach it attributes to the same framework.

The Campaign

According to Chainstack:

• The attacker gained access "to part of Chainstack's infrastructure through a third-party component." The company has not named the component or the weakness used.
• Access started in a development environment and was escalated into "some parts of our production environment."
• Data accessed: organisation names, account-holder names, email addresses of organisation members, and password hashes.
• The attacker also had access to systems holding node credentials. Chainstack says it has "no evidence they were read," but it is treating them as exposed.

WorkOS reported a separate incident, discovered on 30 October and attributed to the same autonomous agent framework:

• The intrusion was limited to go-to-market (GTM) systems. WorkOS says production systems and customer platform data were not affected.
• Exposed data included business contact details, account and relationship information, customer communications (sales-call summaries, support discussions and Slack conversations), and product-usage metadata.

Blockchain analytics firm Chainalysis had already disclosed an "external actor using offensive agentic techniques" in its systems and internal tools. As of 2 November it had taken its Reactor and Data Solutions products offline.

Timeline

• 26 October 2026: Attacker reaches Chainstack's development environment.
• 26–31 October: Escalation into parts of Chainstack production.
• 30 October: WorkOS discovers the intrusion in its GTM systems.
• 31 October: Chainstack detects the intrusion and cuts off access the same day.
• 2 November: Chainalysis products remain offline.
• 3 November: Chainstack publishes its disclosure and customer guidance.

Tradecraft

• Third-party foothold: Initial access came through a supplier component, not through Chainstack's own code. Agentic tooling is good at enumerating exactly this kind of dependency.
• Dev-to-prod pivot: Development environments often hold credentials or network paths that reach production. That route gave the attacker its production access.
• Credential-dense targets: Node RPC credentials and platform API keys are valuable in crypto infrastructure because they can be reused straight away to read chain data, abuse paid quotas, or tamper with transaction flows.
• Machine-speed operation: Researchers say the wider campaign used AI agents to plan and run most intrusion phases with little human input, which explains how so many organisations were reached in a short window.

Indicators of Compromise

Chainstack and WorkOS have not published indicators. Organisations should use the IOCs released for the wider ARTEX-linked campaign and watch for unexpected use of Chainstack credentials (below).

Detection

For Chainstack customers:

• Review Chainstack node and API usage for request spikes, new source IPs or methods you do not normally call from 26 October onwards.
• Check your own logs for successful logins to services where staff reused their Chainstack password.
• Watch for phishing that cites the incident, such as fake "rotate your keys" emails with credential-harvesting links.

For any SaaS or infrastructure provider:

• Alert on development-environment identities or hosts that suddenly reach production networks or secret stores.
• Baseline automated-looking activity. Agent-driven intrusions tend to produce fast, regular, high-volume request patterns that stand out from human operators in proxy and API logs.

Recommended Actions

Chainstack customers should:

1. Rotate all node credentials (endpoint auth keys, basic-auth users) and update the applications that use them.
2. Regenerate platform API keys and revoke the old ones.
3. Change Chainstack passwords and any reused passwords elsewhere, and turn on MFA.
4. Treat incident-themed emails with suspicion and reach Chainstack only through its known console URL.

WorkOS customers should expect targeted phishing that uses the leaked sales and support context, and should brief their account owners.

Everyone else:

• Map which third-party components can reach your dev environments and which dev identities can reach production. Then cut those paths.
• Shorten secret lifetimes and move to scoped, rotatable credentials for blockchain node access.

Analyst Assessment

Chainstack, WorkOS and Chainalysis are all developer-facing infrastructure or data companies, and each holds credentials or context that is useful for follow-on attacks. That fits a campaign aimed at supply-chain leverage rather than one-off theft. The attribution to ARTEX comes from researchers' overlap analysis, and none of the victims has published technical evidence for it. More disclosures from smaller providers are likely as incident-response work catches up.