Retailers Gain AI Control as Agent Risks Spread
Approved tools now dominate, but customer data and remote AI connections remain difficult to monitor.
Retail businesses are bringing employee use of artificial intelligence under tighter central oversight. Netskope found that 73% of retail workers used company-approved AI tools, up from 40% in its earlier report. Use of unsanctioned tools dropped to 44% from 70%. About two-thirds of employees used a separate AI application.
The broader risk is harder to see. Nearly every retail worker used software with built-in AI, and 90% used systems trained with customer information. Of the policy violations Netskope recorded, 56% involved information covered by federal or state rules. Source code accounted for 20%, while passwords or API keys made up 16%.
Autonomous AI systems created another oversight problem. Netskope recorded a 400% rise in retail AI agents connecting with remote Model Context Protocol servers, which can create new routes into outside systems. Phishing campaigns impersonating familiar AI products also returned in early 2026, reaching about 100 users per 100,000 by March. The report covers data gathered July 1, 2025, through July 30, 2026. Netskope recommends restricting unnecessary applications, reviewing web traffic and applying controls that flag sensitive data sent to unapproved tools.
- Approved AI use reached 73% among retail employees.
- Customer-data-trained AI tools were used by 90% of workers.
- Remote MCP connections involving retail AI agents rose 400%.
- The report covers July 1, 2025, through July 30, 2026.
Summarized from the reporting above. Read it for the full story.
A one-page calendar of the federal dates that matter most this quarter, so nothing shows up as a surprise.
Talk through your books, a notice you got, or what's due next quarter. No cost, no pressure.
Book a check-in →
Share this story