“The Microsoft Copilot application has been deemed by the Office of Cybersecurity to be a risk to users due to the threat of leaking House data to non-House approved cloud services,” Szpindor’s workplace mentioned. Data leaks have been an issue for generative AI instruments ever since OpenAI’s ChatGPT arrived. In May final yr, The Wall Street Journal reported that Apple warned workers in opposition to utilizing instruments like GitHub’s Copilot assistant for coding and ChatGPT. After uncovering knowledge leak trails, Samsung additionally prohibited staffers from utilizing ChatGPT.

A number of months later in June, San Francisco-based Robust Intelligence consultants demonstrated how Nvidia’s NeMo Framework AI software program will be tricked into revealing personal info and skipping previous its security measures. The U.S. Federal Trade Commission additionally investigated ChatGPT for placing knowledge safety in danger. In November final yr, consultants at Northwestern University unearthed strategies to trick customized GPTs into revealing confidential info.

A month later, leaked paperwork obtained by The Platformer revealed the hallucination issues with Amazon’s Q chatbot and that it was leaking confidential knowledge equivalent to the placement of knowledge facilities, unreleased options, and low cost packages. In one other incident, ChatGPT spilled personal details equivalent to person banking info, forcing the corporate to briefly take it offline. With such a shaky historical past, it is no shock that Microsoft’s Copilot has been banned from official machines, though Congressional staffers cannot be stopped from utilizing these instruments on their private units.