Subscribe

Employees may be feeding confidential business data into the public domain via ChatGPT

Compliance specialists, Legal Island have issued a stark warning to employers that staff could be feeding confidential business data straight into the public domain via ChatGPT.

Before using ChatGPT, users can disable its training mode, a setting that, when left on, allows OpenAI to store and use input data to refine future responses.

It became apparent to Legal Island that many employers are allowing the use of ChatGPT without any proper training and without making their employees aware of the importance of turning off the training function before using the tool. 

A new survey of 100+ organisations, conducted by Legal Island, found that just 4% of users knew how to disable ChatGPT’s training function, the simple privacy toggle that prevents OpenAI from reusing sensitive data. 

Barry Phillips, Chairman of Legal Island and author of a new book, ‘ChatGPT in HR’, put it bluntly: “When the training feature is left switched on, OpenAI can capture the information entered into ChatGPT and recycle it to improve future outputs.

Business Bulletin

“If your staff are using ChatGPT with the training function left on, you’re potentially leaking commercially sensitive data into a giant AI engine.

“That data could pop up in someone else’s prompt next week. It’s a legal, reputational, and regulatory mess waiting to happen.”

Phillips continued: “While it’s encouraging that employees are embracing ChatGPT and teaching themselves how to use it, the lack of formal training is alarming.

“Our research shows a worrying knowledge gap as most employees in Ireland don’t even know the tool has a training function, let alone how to disable it.”

To address this compliance crisis, Legal Island has developed a free 10-minute e-learning module for employees, showing them exactly how to use ChatGPT safely, including how to switch off the training function before entering any data.

confidential business data
A new survey of 100+ organisations, conducted by Legal Island, found that just 4% of users knew how to disable ChatGPT’s training function. (Photo by Thomas Trutschel/Photothek via Getty Images)

Kellie Shields, Chief Compliance Officer at Legal Island, added: “People treat GenAI like a harmless toy, it’s anything but.

“Without proper training, it’s a data breach in the making. This issue is too important to ignore, so we’re encouraging employers to take action today and avail of the free compliance training.”

Legal Island has been guiding HR professionals through the complexities of workplace compliance since 1998.

Photo: Barry Phillips, Chairman of Legal Island.

Sign up to The Business Plus Panel to help shape the business decisions of tomorrow and win vouchers for your opinions! 
chevron-down linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram