Chainstack Detects Breach on Oct. 31, Links It to Wider AI-Backed Campaign
Key Facts
Chainstack said a breach affecting part of its infrastructure was linked to a wider AI-backed attack campaign, and that it detected the intrusion on Oct. 31, 2026, and cut off access paths the same day. The company said the attacker entered through a third-party component, then accessed parts of its production environment. Chainstack concluded its investigation and closed the incident on Nov. 2, 2026. It said in a statement: "Our findings indicate that this incident was part of a wider AI-backed attack campaign reported in late October." The breach places the platform among a broader wave of activity linked to AI tools, without the company identifying who carried it out.
According to the timeline published by the company, access began in a development environment on Oct. 26, before extending to parts of the production environment on Oct. 29. On Oct. 31, Chainstack detected the breach and closed the entry points, then concluded its investigation two days later. The company said its customer-facing services continued to operate normally throughout the incident. It also said customer nodes, the console and the API were not modified. These points distinguish access to the platform's systems from changes to customer services or interfaces, which Chainstack said did not occur.
Information that may have been exposed could include organization and account-holder names, email addresses of organization members registered with Chainstack, and password hashes. The company said the attacker also accessed systems holding node credentials, which organizations use to connect to nodes they operate through the platform. But it said it had no evidence that the attacker read those credentials, although it was treating them as exposed. It wrote: "We have no evidence they were read, but we are treating them as exposed and asking customers to rotate them." The company therefore asked customers to change passwords, rotate node credentials and reissue platform API keys if they use them.
The technical details published said the initial entry point was a test version of Grafana in a development cluster. The version had been deployed in July and was accessible online without an additional protection layer such as single sign-on or basic authentication, while the default administrator credentials had not been changed. Chainstack said the attacker used ARTEX, an open-source framework that coordinates language-model agents to automate intrusions. It added that it found no evidence of a software vulnerability being exploited, and attributed the breach to the collection of default credentials, broad permissions and exposed administrative functions. The company enabled customers to rotate node credentials themselves through the console, and said: "Rotate your node credentials; self-service rotation is now available in the console."
Chainstack did not attribute the breach to any particular actor, despite linking the incident to the wider campaign through the ARTEX tool, naming indicators and some shared techniques. It said public reports of different groups using the framework do not identify those behind this intrusion or prove that the same operators carried out other incidents. Separately, CrowdStrike said a campaign targeting financial institutions in South Korea from late September to early October 2026 resulted in data exfiltration, without naming the actor behind it. CrowdStrike assessed with medium confidence that those behind the activity were Chinese-speaking and financially motivated, while stressing that it had not attributed it to a specific adversary. Chainstack also said it does not store payment-method details such as full card numbers or security codes, so that data was not exposed in the incident.