Microsoft briefly limited employee access to OpenAI’s ChatGPT, citing security concerns

Microsoft has invested billions of dollars in OpenAI. But on Thursday, employees of the software company were briefly banned from using the startup’s best-known product, ChatGPT, CNBC has learned.

“Due to security and data issues, some AI tools are no longer available to employees,” Microsoft said in an update on an internal website. CNBC also reviewed a screenshot showing that ChatGPT was not accessible on corporate devices.

Representatives for Microsoft and OpenAI did not immediately respond to requests for comment.

“While it is true that Microsoft has invested in OpenAI and that ChatGPT has built-in protections to prevent improper use, the website is nevertheless an external third-party service,” Microsoft said. “That means you have to be careful with it because of the risks to privacy and security. This also applies to other third-party AI services, such as Midjourney or Replika.”

The company initially said it was banning ChatGPT and the design software Canva, but later removed a line in the advisory that included those products. After the initial publication of this story, Microsoft restored access to ChatGPT.

In a statement to CNBC, Microsoft said the temporary blocking of ChatGPT was an error resulting from a test of large language model systems.

“We were testing endpoint control systems for LLMs and accidentally enabled them for all employees,” a spokesperson said. “We restored service shortly after we discovered our error. As we have said previously, we encourage employees and customers to use services like Bing Chat Enterprise and ChatGPT Enterprise that offer a higher level of privacy and security protection.”

Many large companies have restricted the use of ChatGPT, often to prevent the sharing of confidential data. ChatGPT is trained on extensive Internet data and crafts human-like responses to people’s chat messages. The service has more than 100 million users.

Microsoft’s update recommends people use the company’s own Bing Chat tool, which relies on OpenAI artificial intelligence models. The two companies are closely linked. Microsoft has also been busy this year releasing updates to its Windows operating system and Office applications that take advantage of OpenAI services, which in turn run on Microsoft’s Azure cloud infrastructure.

Earlier this week, CEO Satya Nadella appeared on stage alongside OpenAI’s Sam Altman at the startup’s first developer conference.

Altman wrote in one after on X late Thursday that “the rumors that we are blocking Microsoft 365 in retaliation are completely baseless.”

In January, a senior Microsoft engineer wrote on a forum that employees could use ChatGPT, but advised against entering confidential information, Insider reported.

Earlier this week, a hacking group called Anonymous Sudan said it was targeting ChatGPT in an attack because of “OpenAI’s cooperation with the occupation state of Israel” and because Altman said he is “willing to invest more in Israel.”

WATCH: OpenAI on track for $1.3 billion in revenue by 2023, up 4,543% from last year