Back Chosun

Infostealer Malware Now Targets AI Accounts

Chosun • September 20, 2026

Malware steals session cookies, accessing corporate data and AI resources without passwords

The infostealer malware, which previously stole users’ passwords, credit card information, and cryptocurrency wallet data, has now set its sights on generative AI (artificial intelligence) accounts. This shift occurs as AI services accumulate sensitive information such as work documents and source code, while also offering access to high-value AI computational resources.

According to a September 18 report by Japan’s Nikkei newspaper, Israeli cybersecurity firm KELA confirmed that 49,700 session cookies (authentication tokens) stolen from AI platforms circulated on the dark web between January and July this year. Many of these cookies could still be used for active logins. Session cookies are issued to users after they authenticate via passwords and two-factor verification. By stealing these, attackers can access accounts without knowing the password—a tactic akin to stealing a user’s “access pass” instead of breaching the “main gate” of authentication.

AI accounts have become attractive targets for hackers due to their potential to compromise both personal and corporate data in one go. General users often interests, schedules, and photos with AI, while employees analyze internal documents, business strategies, and financial data. Developers, meanwhile, rely on AI for writing and debugging source code.

In March, Microsoft revealed that approximately 900,000 cases of malicious browser extensions collecting ChatGPT and DeepSeek conversation data were installed, with related activities detected in over 20,000 enterprise and institutional work accounts. A recent analysis by U.S. authentication and security firm Octa found that among 44,791 authentication tokens extracted from 5,871 infected PCs across 162 countries, 555 were linked to AI services. API keys for Google Gemini and OpenAI were also discovered.

AI usage permissions themselves are now targets. Anthropic disclosed cases where hackers stole corporate AI API keys to exploit victims’ resources for their own AI usage and attacks on other businesses. “When criminals steal AI authentication data, they gain three advantages: sellable assets, computational resources at the victim’s expense, and the ability to impersonate the victim,” Anthropic stated.

로그인을 하시면 모든 기사를 읽을 수 있습니다

The full story

This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.

Infostealer Malware Targets AI Accounts with Session Cookie Theft Infostealer malware has evolved to target generative AI accounts, stealing session cookies to bypass traditional authentication methods like passwords and two-factor verification. A report by KELA revealed that over 49,700 session cookies from AI platforms were sold on the dark web from January to July 2026, many of which remain valid for logins. This malware not only compromises personal data but also corporate resources, as employees use AI for sensitive tasks. Microsoft previously identified 900,000 installations of malicious browser extensions that collected sensitive AI conversation data. A recent analysis by Octa found 555 authentication tokens linked to AI services among 44,791 tokens extracted from infected PCs. The theft of API keys for platforms like Google Gemini and OpenAI has also been confirmed, allowing attackers to exploit AI services at the victims' expense. The situation highlights the emerging criminal ecosystem around stolen AI credentials. 2 sources 63

Extracted Entities

Infostealer Malware Now Targets AI Accounts

The full story

Extracted Entities

Attack Types (1)

Companies (1)

Countries (1)

MITRE ATT&CK (1)

Platforms (2)

Tools (1)