Infostealer Malware Targeted Generative AI Accounts
Malicious extensions and cookie theft compromised thousands of corporate AI accounts throughout early 2026.
Updated on Sept. 20, 2026 in Cybersecurity
Live Poll
Do you trust the security of your personal data when using generative AI tools?
Between January and July 2026, cybercriminals circulated over 49,700 stolen session cookies from AI platforms on dark web markets. Research into 7GB of infostealer data revealed widespread unauthorized access across 162 countries.
Why it matters
AI accounts have become high-value targets because they often house sensitive internal business strategies, source code, and user documents. Criminals also exploit stolen API keys to leverage expensive computing resources at the victim's expense.
A security analysis of 7GB of infostealer data uncovered 555 authentication tokens linked to AI services out of 44,791 total tokens reviewed. This follows a March 2026 report identifying 900,000 malicious browser extension installations.
The players
Microsoft
A global technology company that develops the Windows operating system, Azure cloud infrastructure, and enterprise-grade security software.
Anthropic
An AI research and deployment company that focuses on building reliable, steerable, and interpretable AI systems.
Octa
A cybersecurity firm specializing in threat intelligence analysis and monitoring dark web data exfiltration.
The details
Infostealer malware functions by infiltrating personal computers to scrape stored passwords and session cookies—small files used by websites to remember login states—directly from the browser's memory. By stealing these active cookies, attackers can bypass multi-factor authentication (MFA) protocols entirely, as the server treats the attacker's device as an already-authenticated session. In other instances, attackers deployed malicious browser extensions to collect conversation data from over 20,000 corporate and institutional accounts.
Timeline
January 2026 to July 2026: More than 49,700 AI session cookies were traded on dark web marketplaces.
March 2026: Microsoft identified 900,000 installations of browser extensions used to collect AI data.
The Tech Race
This wave of account theft highlights the security gap as AI evolves from simple chat interfaces into persistent agents performing complex human tasks. Protecting these credentials is now a primary front in the race to secure enterprise-grade automation.
Users can reduce risk by removing unused or unverified browser extensions and clearing active browser sessions regularly. Companies should implement strict endpoint monitoring to detect malicious software that attempts to extract authentication tokens from browser memory.
The takeaway
As generative AI shifts toward executing autonomous workflows, session-based hijacking will likely become a more frequent vector for corporate espionage. Security teams should prioritize monitoring for anomalous API usage and unauthorized browser extension activity to mitigate these persistent risks.
Further reading
For more on evolving threat landscapes, visit Cybersecurity.
Live Poll
Do you trust the security of your personal data when using generative AI tools?
