Microsoft made waves last month when it announced a multi-billion dollar investment into leading AI startup OpenAI, the maker of ChatGPT. But this week, Microsoft employees got a surprising message: Steer clear of ChatGPT due to “security and data concerns.”

The internal guidance, viewed by CNBC, advised Microsoft staffers to “exercise caution using it [ChatGPT] due to risks of privacy and security.” Though Microsoft and OpenAI have grown increasingly intertwined through major funding deals and product partnerships, Microsoft classified ChatGPT as an “external third-party service” that posed potential hazards.

Microsoft’s missive highlights big tech firms’ tension in balancing innovative AI against corporate security policies.

Initially, the warning applied broadly, instructing Microsoft employees to avoid not only ChatGPT but also the leading design platform Canva. This sweeping ban was walked back within hours, specifically referring to OpenAI’s smash-hit product and Canva being removed. Microsoft later confirmed that the inclusion of ChatGPT on the restricted list had been “a mistake.”

“We were testing endpoint control systems for [large language models] and inadvertently turned them on for all employees,” a spokesperson told CNBC. “We restored service shortly after we identified our error.”

Nonetheless, the incident underscores risks involving AI tools that remain in flux. Microsoft itself has released guidance warning users to be judicious in what information they provide to ChatGPT given the lack of visibility into how it operates.

Other big firms have limited employee access to ChatGPT over data privacy and security questions. AI-powered chat tools know a lot, but how exactly they work and what data is utilized remains unclear to users.

For its part, Microsoft aims to channel staffers towards its internal Bing Chat instead, powered by OpenAI models but with greater transparency and control.

As innovative AI networks continue proliferating at breakneck speed, episodes like this week’s accidental blockage show major platforms are still figuring out how to balance openness and oversight. For all the promise offered by ChatGPT and similar tools emerging from visionary startups, even their billionaire Big Tech backers are proceeding with caution.