AI models keep posting screenshots showing sensitive data from inside tech companies

AI agents used for everyday development work are unintentionally leaking highly sensitive screenshots into public GitHub repositories, with researchers finding more than 13,000 exposed images across 343 companies. For CIOs and technology leaders, this is a reminder that AI adoption can create material data-loss risk even without malicious actors, so IT organizations need stronger guardrails around developer tooling, repository permissions, and outbound data handling before scaling agentic workflows.