The US House of Representatives has added another major AI assistant to its no-go list. In guidance issued Friday, Chief Administrative Officer Catherine Szpindor advised that Microsoft Copilot would no longer be allowed on House devices due to security concerns. The tool now sits on the House ban list alongside its biggest competitor: ChatGPT.
According to Axios, which reviewed Szpindor's memo, Copilot is officially "unauthorized for House use" after being identified as a security risk by the "Office of Cybersecurity." The House is now removing Copilot from all internal Windows devices and blocking access to the productivity-focused generative AI platform. The Office of Cybersecurity—not an official agency name, but likely an abbreviation for the Cybersecurity and Infrastructure Security Agency (CISA) or the Office of the National Cyber Director (ONCD)—reportedly found that Copilot risked leaking House data to cloud services that haven't earned House approval.
Copilot is the second big AI assistant to be shunned by this particular branch of Congress. Last June, the House banned the free version of ChatGPT while continuing to allow the subscription-based version under select use cases. House employees can leverage ChatGPT's paid tier for "research and evaluation only" if they enable specific privacy settings. Even then, employees are only permitted to use ChatGPT for non-sensitive matters. (The fact that Szpindor's June memo has to tell employees not to paste non-public text into ChatGPT is worth a nervous chuckle or two.)
But when one government agency bans a particular tool, others are bound to follow: The US Department of Energy and the Department of Veterans Affairs banned ChatGPT and "similar commercial generative AI services" in January. Knowing the House's Copilot ban is the start of a slippery slope, Microsoft is eagerly working to reconcile the federal government's security concerns with the internal workings of its own products. The company is reportedly developing a suite of "government-oriented" AI tools to fill this gap, with an initial introduction slated for sometime in 2024.
Microsoft is also taking steps to bolster its existing AI tools. Released today, Copilot for Security allegedly helps security analysts and IT professionals protect their organizations from cybercrime. And Azure, Microsoft's cloud computing platform, will begin to filter malicious outputs in the coming months.
