More Companies Are Banning ChatGPT Due to Privacy Risks

A growing list of large companies are banning the use of ChatGPT by employees, at least for the time being, citing privacy risks.

Samsung reported a leak of confidential source code after employees used OpenAI’s ChatGPT chatbot, and Apple recently barred the use of ChatGPT and other third-party AI programs.

Bank of America, Goldman Sachs, JPMorgan Chase and other banks have restricted the use of ChatGPT by employees. Defense contractor Northrop Grumman and telecommunications company Verizon have also blocked ChatGPT.

Companies are restricting the use of ChatGPTand other third-party generative AI tools “because there are concerns of personal or company data being shared to a company like ChatGPT that can use it for other purposes,” said Jodi Daniels, founder and CEO of privacy consulting firm Red Clover Advisors. "There are concerns that company confidential data can be shared without permission by its current customers.”

ChatGPT uses data from its interactions with users to improve its accuracy, though users can opt out of this in the settings.

But in addition to privacy concerns, there are other risks involved with using ChatGPT in the workplace, Daniels said.

"ChatGPT scours the internet and summarizes existing data. It has notoriously been wrong and the information provided back may not be allowed to be used due to copyright restrictions,” Daniels said. “If a person of company does not fact check the information or change it into their own words, they could be infringing on someone else’s work and plagiarizing.”

Some companies currently restricting the technology have said they’re open to using it after it’s  more thoroughly vetted, while others have opted instead to develop their own internal AI tools. Whichever route a company chooses, clear policies around the use of AI tools can help protect companies and their employees, Daniels said.

“Companies should have an AI policy that describes what employees can and can’t use AI tools for, what type of data can and can’t be shared, and what type of business use cases are okay or not okay to use,” Daniels said.