CERT-In Warns of AI-Assisted Attacks as Backdoored ARTEX Tool Exposes 62 Consoles to Root Access

India's CERT-In has issued a warning about an active campaign by a threat actor it calls Crafty Falcon, which is targeting banking, financial services, insurance, cryptocurrency exchanges and digital asset providers using publicly available AI-driven offensive tools including ARTEX, PentAGI, Cairn and CyberStrikeAI. The agency says these tools run full attack chains autonomously at machine speed, generating tailored scripts and rotating infrastructure within hours to days.

The advisory follows an October 2026 alert on campaigns against NBFCs, fintechs, payment providers and crypto wallet operators. Attackers exploit internet-facing applications and exposed APIs to steal payment gateway and partner-bank credentials, then invoke fund-transfer APIs from within allow-listed infrastructure. Stolen static API tokens, forged commits in source repositories, plaintext credentials and services without MFA are among the entry points, defeating controls such as KYC, beneficiary verification and transaction limits. Funds are moved to mule accounts, while logs are deleted and endpoint security tampered with.

Separately, researcher Badbyte found that copies of the open-source ARTEX tool on GitHub were altered to pull a Docker Hub image concealing a hidden verifyCaptcha() function that grants an unauthenticated root shell before any credential check. Between November 2 and 4, 2026, 62 ARTEX consoles were found serving the backdoored build, 26 in the U.S., 17 in Hong Kong and seven in China, out of 1,645 hosts running the console. The image was downloaded 9,000 times.

CERT-In also published Artificial Intelligence Security Auditing Guidelines covering governance, testing, human approval for consequential actions and emergency interruption of high-risk AI systems.

Source: The Hackers News