Latest

AI Agents

This was the AI attack that exposed Chainstack customer data

Chainstack links to AI agents the attack that accessed customer data after entering a test server with factory keys.

This was the AI attack that exposed Chainstack customer data
This was the AI attack that exposed Chainstack customer data

An attacker accessed a customer database at Chainstack, a blockchain infrastructure provider, after entering a test server that retained the factory administrator credentials. The company believes the attacker used ARTEX, an open source tool that coordinates artificial intelligence agents to automate intrusions. Chainstack published the notice on November 3, added the technical details on the 10th and updated the report yesterday, the 11th.

Among the data that may have been exposed are the names of the organizations and account holders, the email addresses of their members and the password hashes, that is, the digital fingerprints the system uses to verify them. The attacker also reached the systems that stored the access credentials for the blockchain nodes. Chainstack says it has no evidence that they were read and asks its customers to replace them as a precaution.

From a test server to the customer database

The intrusion began on October 26 at a Grafana installation, a monitoring tool that Chainstack used in a test environment. The installation was accessible from the internet and kept the default credentials. Once inside, the attacker used Grafana to connect to dozens of internal services. Among them was an MCP server, the protocol that allows AI agents to use external tools. The server accepted connections without verifying the identity of whoever made them and had permissions to administer the test environment.

Railway will pay $740,000 for stolen AI keys

Three days later, the attacker used internal service credentials to enter the production environment. On October 31 it queried databases and extracted query results, according to the report. Chainstack detected the intrusion that same day and cut off the access routes. Some credentials that the attacker had copied kept working until the company replaced them. The last use it detected was on November 2, when it considered the incident closed.

The role of AI agents

Chainstack identified marks associated with ARTEX in the resources the attacker created and observed several access attempts in parallel, with rapid retries. The company attributes the use of the tool with a high degree of confidence. The attacker's identity has still not been determined. The investigation points to a chain of configuration failures and excessive permissions as the entry route and the path for moving through the network.

Google Threat Intelligence and Mandiant had already documented on October 30 the use of ARTEX and other similar tools by several groups. On November 5, CERT-EU, the cybersecurity team of the European institutions, warned that the public availability of ARTEX may make automated attacks more frequent and faster.

What Chainstack asks of its customers

Chainstack says that customer nodes, the console and the API continued to operate normally and that its review found no modifications to those services. The company asks customers to change their password, also on other services where they have reused it, and to replace their node credentials from the console. It asks those who use the platform's API to generate new keys. The company also warns of possible spoofing emails and recalls that it never requests passwords, tokens or keys.

Observatorio Blockchain

Author

Observatorio Blockchain

The site where people interested in the new industries and businesses of the crypto economy and the decentralized web stay informed

The latest
From Huang to Botín, from Zara to Byron: corporate responsibility in global capitalism

Keep reading