Microsoft has invested billions of dollars into AI startup OpenAI, becoming one of their biggest backers.
However, Microsoft employees recently found themselves blocked from accessing OpenAI’s wildly popular ChatGPT tool on corporate devices, as Microsoft cited “security and data concerns.”
According to an internal update viewed by CNBC, Microsoft informed employees that “a number of AI tools are no longer available for employees to use“, with ChatGPT specifically called out as being blocked.
The tech giant initially also included design platform Canva in their banned services list, but later removed this reference.
After the story first broke, Microsoft moved swiftly to restore access to ChatGPT, claiming the block had been an error resulting from testing of systems for large language models. A spokesperson clarified that
“We were testing endpoint control systems for LLMs and inadvertently turned them on for all employees. We restored service shortly after we identified our error.”
However, the incident has raised eyebrows about what security flaws Microsoft may have uncovered in ChatGPT that have not yet been made public.
With over 100 million users interacting with the AI chatbot, any vulnerabilities could have far-reaching implications.
ChatGPT composes remarkably human-like text in response to user prompts, having been trained on a massive trove of internet data. But this data exposure also carries risks, as the AI could share sensitive information or be tricked into generating harmful content.
While Microsoft is continuing to integrate OpenAI tools like ChatGPT into products like Windows and Office, they appear to be treading carefully.
The company advised employees to instead use Microsoft’s Bing Chat for now, which also utilizes OpenAI models but comes with “greater levels of privacy and security protections.“
Industry experts note that many big corporations are similarly cautious about unfettered access to platforms like ChatGPT, afraid that confidential data could be leaked or company secrets revealed.
But they say more transparency is needed from Microsoft and OpenAI about what exact security flaws exist and how they are being addressed.
As AI services become more advanced and ubiquitous, pressure will grow on developers like OpenAI to restrict potential harms without limiting beneficial uses.
For now, Microsoft’s security team seems to have sounded the alarm on ChatGPT’s dark side, but the details remain obscured.
