Enterprises Build 4-Part Defenses Against Shadow AI as 67% Use Non-Corporate Accounts
Updated
Updated · Computerworld · Oct 19
Enterprises Build 4-Part Defenses Against Shadow AI as 67% Use Non-Corporate Accounts
3 articles · Updated · Computerworld · Oct 19
Summary
67% of employees access AI services through non-corporate accounts on work devices, pushing enterprises to tighten controls around “shadow AI” as agentic tools can now take risky actions, not just leak pasted data.
A four-part response is emerging: provide sanctioned AI alternatives, monitor and block unsafe use, train employees on policy, and impose consequences when they deliberately bypass controls.
BitGo and Docusign said they pair approved enterprise versions of ChatGPT, Claude, Gemini and Grok with DLP, network controls and blocks on tools such as DeepSeek, while treating agentic frameworks like OpenClaw as high-risk.
1.9 million AI session minutes analyzed by Harmonic found 65% of personal-account activity was work-related, and a separate report said every monitored organization had at least one employee sending data to a China-based AI tool.
66% of employees have used unsanctioned AI, according to a PagerDuty-backed survey, and experts say weak enforcement could leave companies legally exposed if AI agents harm outside parties.