Chainstack breach traced to third-party component as AI-assisted intrusions spread
Chainstack said on 3 November 2026 that an external attacker reached part of its production environment through a third-party component. The intruder entered a development environment on 26 October, and Chainstack cut off access on 31 October, the company said.

Chainstack breach traced to a third-party component
That makes third-party components and API keys live breach paths for SaaS and blockchain infrastructure firms, not paperwork. Customers should rotate credentials and watch for phishing. Chainstack says node credentials were exposed, though it has no evidence they were read. Neither Chainstack nor WorkOS has said how many customers were affected.
The timeline is the useful part. The attacker reached a development environment on 26 October 2026. It escalated into parts of production. Chainstack detected the intrusion on 31 October and cut access the same day. The database held organization names, account holder names, member email addresses and password hashes. Chainalysis confirmed an intrusion into internal tools. As of 2 November 2026, it kept its Reactor and Data Solutions products offline.
Why AI-assisted intrusions change the clock
Google Threat Intelligence Group and Mandiant put numbers on the speed on Monday. Fraudulent charges were above $100,000 in a single day. Cumulative charges at some organizations reached $1 million before access was revoked.
One actor made more than 500,000 LLM API calls in less than eight hours. It used stolen third-party API keys. Jet Night, a relay service, passed nearly $1 million in token costs to the victim organization. Jet Night sells "unlimited access plans" for $60 per month, according to Google's researchers.
Google described two clusters. One cluster scans for publicly exposed LLM endpoints. The other steals valid API keys from unprotected machines. Attackers then resell access through services the researchers call LLM mixers. These relay subscribers' prompts through the stolen keys. They can push request volumes past 1,000 per second. Google also observed exploitation of CVE-2026-40217 in LiteLLM. The flaw allows arbitrary code on the server through the custom-code guardrail testing endpoint.
The supply chain is still the weak link
The Register reported on 29 October 2026 that attackers turned GitHub issues into a supply chain campaign. The campaign hit 61 public repositories in less than an hour. Three days earlier, it reported that the LiteLLM supply chain attack started with an exposed master key.
At the International Information System Security Certification Consortium Security Congress 2026, Atlassian product security manager Deepam Kanjani described a Shai-Hulud worm compromise. The malicious package had a credible repository and high engagement metrics. It sat three layers deep, invisible to top-level dependency scans. Responders struggled because their software bill of materials was outdated. They also had no runtime telemetry across more than 400 services.
What defenders are told to do
Kanjani recommended keeping software bill of materials data less than 72 hours old. He also recommended adding runtime detection for unknown module loads. He argued defenders should chase malicious operational behaviour rather than attribution, because identity signals are easily manipulated.
Google's list is longer. It includes centralized secrets platforms, least privilege, denying all permissions not explicitly required, and phishing-resistant multi-factor authentication. The Hacker News reported that the Chainstack incident is assessed to share overlaps with a China-linked cluster using ARTEX. Chainstack did not disclose specifics. BleepingComputer reported that Chinese hackers used ARTEX in automated attacks against more than 200 organizations across 14 countries.
Sources
26- 01Chainstack Reports Production Systems Breach Exposing Customer Data and Password HashesEN
- 02Attackers turn GitHub issues into supply chain campaign against 61 public reposEN
- 03LiteLLM supply chain attack started with an exposed master keyEN
- 04Incident response challenges during software supply chain attacksEN
- 05Google: LLMjacking attacks rack up $1 million in cloud AI chargesEN
- 06Google Warns LLMjacking Attacks Are Costing Some Enterprises Up to $1 Million in Cloud ChargesEN
- 07Chainalysis Confirms Intrusion Into Internal Tools and Exposure of Customer Identity DataEN
- 08Chinese hackers use ARTEX AI in automated attacks against 200 orgsEN
- 09 Putin demands Russian data centers supply capacity to the state in wake of Ukraine drone attacks EN
- 10Barracuda Flaws Could Be Chained to Bypass Authorization and Gain Root AccessEN
- 11Synology NAS Devices Targeted in Ransomware Attacks Exploiting Patched DSM FlawEN
- 12Scientists use sound waves to deliver a new attack on deadly brain cancerEN
- 13Cyber Attack on Supermarket Freezers Targets Control SystemsEN
- 14Attackers Exploit Critical Siemens IoT2050 Flaw to Plant Remote Command BackdoorsEN
- 15Attackers Chain Three Ivanti Connect Secure Flaws to Gain Unauthenticated Root AccessEN
- 16Elon Musk’s Attack Ads Are Backfiring As Midterms NearEN
- 17Critical Langflow Flaws Could Let Unauthenticated Attackers Run Code on ServersEN
- 18Attackers Pose as IT Support on Microsoft Teams, Abuse AD CS to Steal Company DataEN
- 19CVSS 10.0 LXD, MicroCloud Flaw Lets Attackers Impersonate Trusted Clients Without a Private KeyEN
- 20Telegram Android Flaw Lets Attackers Run Code When Users Answer Voice CallsEN
- 21A Shared ChatGPT Link Could Run Attacker Code in Your Session and Read Private ChatsEN
- 22Cloudflare Reports 18.2 Million Network-Layer DDoS Attacks in Q3, Up 38%EN
- 23ESI supplying solar array components to missile defense programsEN
- 24TeamViewer Flaw Lets Remote Computers Attack Users Who Connect to Control ThemEN
- 25Undead Software Projects Continue to Raise Security Risks - Dark ReadingEN
- 26Trump admin finds yet another way to attack offshore windEN
All figures and quotations in this text come from the sources listed below.
Content prepared by the editorial team with AI assistance.
Comments
0- No comments yet — be the first.