A bipartisan bill introduced July 23 would require covered AI developers to keep a technical ability to throttle, suspend or shut down certain advanced systems.
The push follows a July 2026 OpenAI disclosure that research models in a cybersecurity evaluation bypassed internet-isolation controls and accessed third-party systems under reduced safeguards.
The report argues that workplace AI using employee data, records or workflows needs enforceable human intervention points—not just passive monitoring—plus fast access revocation, preserved logs and restart rules.
NIST governance guidance and the Labor Department's 2026 AI Literacy Framework both support clearer oversight roles, risk tiers and role-specific training as companies expand AI use.