TL;DR
- Chainstack disclosed a breach exposing customer information and potentially node credentials, while saying customer nodes, its console and API were not modified.
- The attacker entered through an exposed Grafana instance, reached an unauthenticated MCP server with cluster-admin rights and used ARTEX, a framework that orchestrates LLM agents.
- Chainstack found no evidence node credentials were read, but told customers to rotate them, change passwords, regenerate API keys and monitor phishing attempts.
Chainstack has disclosed a security breach that exposed customer account data and potentially node credentials, linking the intrusion to a broader AI-powered attack campaign. In its official incident report, the Web3 infrastructure provider said an external attacker entered through a third-party component and reached production systems after first compromising a development environment. The incident exposed customer information while leaving customer-facing nodes, the console and API unmodified, adding another security warning for Web3 infrastructure operators. Chainstack said payment method details such as full card numbers and security codes were not stored and therefore were not exposed.
AI Tooling Helped Turn Infrastructure Weaknesses Into Production Access
The attacker first accessed an internet-exposed Grafana instance in a development Kubernetes cluster where default administrator credentials had not been changed and additional authentication was absent. Grafana’s data-source proxy then provided a route to internal services, including an unauthenticated MCP server running with cluster-admin privileges. Chainstack said it found no exploited software vulnerability. The intrusion instead chained default credentials, missing authentication, excessive permissions and long-lived credentials, showing how AI-linked security threats can amplify weaknesses already present in infrastructure. Initial access occurred on October 26, followed by production access on October 29.
Chainstack assessed with high confidence that the attacker used ARTEX, an open-source framework that orchestrates LLM agents to automate intrusions. Investigators linked the breach to the broader AI-powered campaign through matching tooling, naming markers and techniques, while stressing that this does not identify the actor. The attacker eventually read production secrets, used a compromised VPN key and queried production databases. Customer organization names, account holder names, member email addresses and password hashes may have been exposed, extending recent concerns around credential exposure across crypto companies. Chainstack detected the intrusion on October 31 and cut the access paths.
Systems holding node credentials were also accessible. Chainstack said it has no evidence those credentials were actually read, but is treating them as exposed and asking customers to rotate them. Customers were also instructed to change passwords, regenerate platform API keys where applicable and remain alert for phishing attempts. Chainstack cut the attacker’s access on October 31 and observed the last use of a copied credential on November 2, while rotating exposed credentials, invalidating login sessions and tightening privileges. The response echoes broader crypto infrastructure security measures focused on credential rotation, authentication and limiting access after breaches.
