Infostealer malware is increasingly exposing enterprise AI accounts, creating risks that extend beyond traditional credential theft by giving attackers access to conversation histories, active sessions, API keys, automation capabilities, and billable resources. SOCRadar analyzed more than one million infostealer records associated with AI services across over 80,000 corporate domains, narrowing its research to 482 major enterprises and identifying 5,434 records linked to 1,500 distinct corporate email addresses; 295 organizations had appeared in stealer logs within the previous 90 days. ChatGPT/OpenAI accounted for the overwhelming majority of exposure, appearing at 358 companies and representing roughly 90% of records, while developer and automation platforms including Hugging Face, Replit, Zapier, and Notion were also affected. Technology and internet services companies recorded the greatest exposure, though financial services, healthcare, energy, retail, and industrial organizations were also impacted. Stolen session cookies are particularly concerning because they can potentially allow attackers to bypass MFA and access sensitive prompt histories, while compromised automation sessions may inherit existing OAuth permissions and stolen API keys can enable unauthorized usage and “LLMjacking.” The findings highlight shadow AI and unmanaged endpoints as significant enterprise risks and recommend organizations discover unauthorized AI accounts, enforce SSO and shorter-lived sessions, rotate and restrict API keys, monitor for anomalous session reuse, and treat employees appearing in infostealer logs as potential endpoint compromises rather than simply resetting their passwords.