Senate offices get ChatGPT, Copilot, and Gemini. The frontier tools at the center of the actual regulatory debate are all off-limits.
ONYX NEWS STAFF — DATELINE: WASHINGTON
| 2-MINUTE CONTEXT: An internal communication obtained by NPR, sent to Senate staff this spring by the chamber’s sergeant at arms, spells out exactly which AI tools are — and aren’t — authorized for use. Most Americans want Congress to regulate AI. The people doing the regulating are largely walled off from the technology itself. |

What’s Actually Authorized
According to the memo NPR reviewed, Senate offices have free access to exactly three AI chat interfaces: Microsoft Copilot Chat, Gemini Chat for Google Workspace Enterprise Plus, and OpenAI ChatGPT Enterprise. The sergeant at arms, who oversees chamber cybersecurity, has not authorized any of the more advanced tools currently at the center of Washington’s AI regulatory debate and the broader public conversation about AI safety.
The stated rationale is a balance between innovation and risk — a reasonable enough institutional posture on its own. The complication is what it means in practice: the people with the most direct power to write AI legislation are, by policy, not using the specific class of tools that legislation is meant to govern.
| “This is a tool that half of all Americans are using today. So if your only mental model of the technology is the harm, you’re probably going to write a bill that’s all brakes and no steering wheel.”— Kovacevich, quoted by NPR |
Not Everyone in Washington Agrees This Is a Problem
One senior Democratic senator, speaking to Axios last week, defended staying hands-off entirely: “you don’t need to do drugs like heroin to know they should be regulated.” Kovacevich’s response to that framing, in the same NPR piece, was pointed: “Well, that’s true, but nobody thinks heroin has an upside to be preserved.”
| THE ACCESS GAP | |
| 3 AI tools authorized for Senate staff use | 0 frontier/advanced AI tools currently authorized |
| ~50% of Americans estimated to be using AI chatbots today | |
| WHY THIS MATTERS: Legislation written by people with no hands-on exposure to the specific tools they’re regulating tends to reflect secondhand impressions rather than direct experience — a gap that shows up later in how workable the resulting rules actually are. |
| WHAT HAPPENS NEXT: No timeline has been announced for expanding authorized tools. ONYX will track whether any Senate office seeks an exception, and whether that request becomes public. |
| SOURCE TRANSPARENCY NPR’s Eric McDaniel — original reporting, direct review of the sergeant at arms’ internal communication, September 23, 2026 |

