Several members of Congress—not known to be a particularly tech-savvy bunch—are trying to get a handle on how to regulate artificial intelligence for the masses, but it appears the crackdown begins at home.
As Axios reports, the Chief Administrative Officer for the House, Catherine L. Szpindor, released a memo that outlines how congressional staffers can use ChatGPT. In short: they can't, unless they're using the $20-per-month ChatGPT Plus.
"The Plus version of the product incorporates important privacy features that are necessary to protect House data," Szpindor writes in the memo.
Large language models (LLMs) like ChatGPT spit out answers to all sorts of queries in seconds, but they also ingest and reuse whatever people type in. Samsung learned that the hard way when three employees used ChatGPT to troubleshoot proprietary code and summarize internal meeting notes, prompting the company to ban its use.
To that end, House staffers are only allowed to use ChatGPT for research and evaluation and can only serve it "non-sensitive data." No asking the chatbot to summarize state secrets.
"House offices are authorized to experiment with the tool on how it may be useful to congressional operations, but offices are not authorized to incorporate it into regular workfow," the memo says.
Furthermore, ChatGPT Plus must be used with privacy settings enabled, which will help to ensure that users' histories are "not preserved" and "interactions are not incorporated back into the massive Language Model for it to learn from."
Sorry, Google Bard. "No other versions of ChalGPT or other large language modelsAl software are authorized for use in the House currently," the House memo concludes.
As Axios notes, members from both sides of the aisle in the Senate and the House are currently working on AI-related legislation, including Senate Majority Leader Chuck Schumer and the unlikely duo of Sens. Richard Blumenthal (D-Conn.) and Josh Hawley (R-Mo.), who presented a framework emphasizing corporate accountability.