A spring‑time internal communication obtained by NPR shows that Senate staff have free access to only three AI chat interfaces – Microsoft Copilot Chat, Google’s Gemini Chat for Workspace Enterprise Plus, and OpenAI’s ChatGPT Enterprise – and no other AI services.
More sophisticated, “agentic” tools such as OpenAI’s Codex or Anthropic’s Claude Code and Cowork, which can autonomously perform tasks and access files, have not been authorized. The sergeant‑at‑arms, who oversees Senate cybersecurity, has kept these capabilities off limits, citing security concerns.
The restriction comes as a majority of Americans call on Congress to regulate AI. Critics argue lawmakers are drafting rules for technology they rarely use. “Congress, unfortunately, has always lagged behind,” said Adam Kovacevich, founder and CEO of the tech‑policy group Chamber of Progress. “Right now you’ve got lawmakers writing rules for technology they, in many cases, never even used.”
The Senate’s policy explicitly forbids any platform from accessing internal drives, shared folders, email, Teams chats or other Senate resources. Daniel Schuman, executive director of the nonpartisan American Governance Institute, warned that “the use of tools…that have the possibility of exfiltrating data elsewhere is a significant risk,” but added that security concerns alone should not block useful technologies.
A Senate staffer who spoke on condition of anonymity said advanced AI tools are being vetted for specific use cases, though no timeline was provided. The staffer noted that security arguments are often used as a pretext to slow adoption in government agencies.
When asked for comment, the top Republican and Democrat on the Senate Rules and Administration Committee declined to discuss details. A committee spokesperson said the policies are designed to “protect Senate data and include stronger security and contractual requirements.”
Across the Capitol, the House of Representatives handles AI oversight through the Office of the Chief Administrative Officer, which permits a broader range of tools from Microsoft, OpenAI, Google and Anthropic under defined use‑case rules.
Despite the Senate’s blanket restrictions, some offices are experimenting independently. Republican Sen. Bernie Moreno of Ohio told a tech‑news show that his office runs an AI “staffer” named Albert for research tasks, and he hopes the Senate will relax its rules next year.
Experts caution that limiting access may leave lawmakers ill‑equipped to understand the technology they are regulating. Kovacevich warned that drafting policy based solely on perceived harms could produce legislation that “has all brakes and no steering wheel.”