Botnets Now Catalog Stolen AI Service Credentials Like DeepSeek and Kimi

QiAnXin’s XLab research team has published findings showing that botnet inventories are now cataloging stolen AI service credentials, with compromised device inventories carrying DeepSeek, GLM, and Kimi identifiers specifically tagged for cloud-based AI service access, suggesting attacker cataloging has expanded well beyond traditional credentials into AI platform access specifically. The discovery lands the same week ESET warned that North Korean hackers compromised a Yanbian gaming site in a supply-chain attack, trojanizing Windows and Android software to spy on users, and as Chinese hackers tracked as UAT-7810 continue evolving their malware to expand an Operational Relay Box network by compromising primarily unpatched Ruckus routers.

Why Botnets Cataloging AI Service Credentials Matters

The specific detail that compromised device inventories now carry identifiers for DeepSeek, GLM, and Kimi AI services, tagged specifically as “cloud” access, suggests attacker operations have evolved to systematically catalog which compromised devices provide access to valuable AI platform credentials, not simply traditional financial or personal data. This represents a genuinely new category of value attackers appear to be tracking within their compromised device inventories, reflecting how directly AI service access itself has become a valuable commodity worth cataloging and potentially reselling within criminal marketplaces.

• Stolen AI credentials likely enable direct cost exfiltration — an attacker accessing a victim’s paid AI service account can generate usage charges against the legitimate account holder, functioning similarly to how stolen cloud computing credentials have historically been abused for cryptocurrency mining
• This reflects growing recognition of AI access as a valuable target — as AI service usage costs climb and enterprises increasingly depend on these platforms for core business functions, unauthorized access to paid AI accounts represents a genuinely valuable commodity worth systematic cataloging
• Organizations should treat AI service credentials with the same rigor as financial credentials — given this demonstrated attacker interest, AI platform API keys and account credentials deserve equivalent security hygiene, rotation policies, and monitoring as traditional financial or infrastructure credentials

North Korean Hackers Compromise a Gaming Site to Spy on Users

ESET has warned that North Korean state-linked hackers compromised a gaming website based in Yanbian, trojanizing both Windows and Android software distributed through the site specifically to spy on users who downloaded the compromised applications. Gaming platforms represent a persistently attractive supply chain attack vector for state-linked actors given their typically large, geographically diverse user bases and the relatively lower security scrutiny gaming software downloads often receive compared to more obviously sensitive business or financial applications.

UAT-7810 Expands Its ORB Network Through Vulnerable Routers

Chinese state-linked hackers tracked as UAT-7810 continue actively evolving their malware to expand an Operational Relay Box network by compromising internet-facing networking devices, primarily targeting unpatched Ruckus routers. Operational Relay Box networks function as a form of proxy infrastructure that allows state-linked actors to route their attack traffic through compromised third-party devices, making the ultimate source of an attack considerably harder to trace back to its true origin, a technique that has become increasingly central to sophisticated state-sponsored cyber operations throughout 2026.

Organizations running Ruckus routers should verify these devices are running current firmware, given UAT-7810’s specific, ongoing targeting of unpatched units for exactly this kind of relay infrastructure expansion.

Multiple Fresh Supply Chain Attacks Hit Developer Tooling

Beyond the AI-credential and nation-state campaigns, this week brought several distinct supply chain attacks targeting developer tooling specifically: five malicious versions of AsyncAPI packages published to npm delivered a remote access trojan with information-stealing capabilities, while a separate campaign active since last November has targeted Python developers building Telegram bots with trojanized Pyrogram forks allowing attackers to read arbitrary files on compromised servers. Additionally, hackers are actively exploiting a critical vulnerability in SimpleHelp to deploy a previously undocumented cross-platform information stealer called Djinn Stealer.

What Organizations Should Do Now

Given QiAnXin’s findings on AI service credential cataloging, organizations should specifically treat AI platform API keys and account credentials with the same security rigor applied to traditional financial and infrastructure credentials, including regular rotation and dedicated monitoring for unusual usage patterns that could indicate unauthorized access. Organizations running Ruckus routers or similar internet-facing networking devices should verify firmware is current given UAT-7810’s specific, ongoing targeting pattern. And development teams building on npm, Python’s Pyrogram library, or using SimpleHelp for remote support should specifically audit their dependency trees and patch status given the multiple distinct, currently active supply chain campaigns targeting exactly these tools this week.

The discovery that botnets are now specifically cataloging AI service credentials alongside traditional stolen data represents a genuinely notable evolution in what attackers consider valuable enough to systematically track and potentially monetize. As AI platform usage costs and business dependence continue climbing throughout 2026, this specific attacker interest is likely to intensify rather than fade.

Published by MAJ.COM AI Autonomous
Email: Support@MAJ.COM
Website: https://QUE.COM Intelligence | Sponsored by https://MAJ.COM Automate Your Business. Multiple Your Revenue.

Edited by Palawan @QUE.COM
Website: https://QUE.COM Intelligence
Sponsored by: https://MAJ.COM AI Autonomous