News Cyber Attacks

AI hacking spree exposed: criminals ran 1,100+ automated attack projects on online stores with the Cairn AI agent platform

Researchers at NorForge found exposed control panels showing at least three threat actors using Cairn, an open-source AI agent platform, to break into e-commerce and payment sites. Reuters reports the largest group breached more than 100 companies.

Published Updated
Illustration of a swarm of glowing AI agent drones attacking a row of online shop storefronts, guided from a dashboard of attack projects.

Key takeaways

  • NorForge found at least three threat actors abusing Cairn, an open-source AI agent platform, to attack websites.
  • The largest, Chinese-language cluster ran over 1,100 attack projects against e-commerce and payment sites.
  • Reuters reported the main actor breached more than 100 companies.
  • The agents aimed for web shells using SQL injection, deserialization, SSRF and other common flaws.
  • NorForge says the activity looks financially motivated, not state-sponsored.

Criminal groups are using AI agents to break into online stores at scale, according to research published on 19 October 2026 by AI security firm NorForge. The attackers used Cairn, an open-source platform that coordinates AI agents toward a goal and can run as an autonomous penetration testing engine.

NorForge found the activity because the attackers left their Cairn control panels exposed on the internet, revealing their targets, instructions and results.

How big is the AI attack campaign? #

NorForge grouped the panels into three clusters:

Cluster Language Scale Targets and goals
AIA0001 Chinese 14 panels, over 1,100 attack projects E-commerce and payment infrastructure in the US, Japan and Australia, seeking persistent access
AIA0002 English 10 panels Credit card and cryptocurrency theft from merchant payment, gray-market and pharmaceutical sites, plus some African commerce sites
AIA0003 English 1 panel, 59 projects Smaller e-commerce sites and large US commercial platforms; appears exploratory

NorForge says the attacks succeeded about 10% of the time on average, with many more reaching partial goals. For the largest cluster, successful projects took under 23 hours on average from start to goal. Reuters reported that the main actor breached more than 100 companies, SecurityWeek said.

How did the AI agents break in? #

According to NorForge, targets were ranked and pre-checked with the Playwright browser automation framework, then handed to Cairn. The agents shared findings on a common board and worked toward a goal, often an upload-based web shell for persistent access. NorForge describes the agents using SQL injection, unsafe deserialization, authentication bypass, server-side request forgery (SSRF) and server-side template injection (SSTI).

When NorForge scanned the material with the TruffleHog secret scanner, it found hundreds of exposed secrets, about 11% of which were verified as valid.

Who is behind it? #

NorForge says the activity appears financially motivated, with no obvious state sponsorship, and that the setup was "almost certainly" criminals reusing existing tools. It says the configuration was repurposed from a capture-the-flag setup and that it has no reason to believe Cairn's original developers are involved. NorForge believes the operator of the second cluster is based in Africa. The research built on a September 2026 report by Gambit Security on AI tools used against online retailers.

NorForge says it is working with Reuters to notify victims and is seeking takedowns of the panels.

What should online businesses do? #

  1. Patch and harden web applications against the flaw types the agents used: SQL injection, deserialization, authentication bypass, SSRF and template injection.
  2. Block or closely monitor file upload paths, since web shells were a common goal.
  3. Scan code, configuration and public assets for exposed secrets, and rotate anything found.
  4. Watch for automated, high-volume probing and unusual browser automation traffic.
  5. Hunt for web shells and unexpected files in web directories if you run e-commerce or payment platforms.

Sources

  1. Inside Cairn: how attackers abused an AI agent platform, NorForge
  2. AI platform Cairn abused to hack over 100 organizations, SecurityWeek
  3. Hackers abuse AI pen-test tool Cairn in attack spree, Dark Reading

Questions this raises

What is Cairn?
An open-source platform that coordinates AI agents toward a goal. It can act as an autonomous penetration testing engine, which attackers repurposed.
Were the attacks successful?
NorForge says about 10% of attack projects reached their goal on average. Reuters reported the main actor breached more than 100 companies.
Are the victims being told?
NorForge says it is working with Reuters to notify victims, especially those that appear compromised.

AI SecurityArtificial IntelligenceAI AgentsCybercrimeCyber AttackThreat IntelligenceE-commerce Security