The Conversation have recently talked about the reality of the data privacy situation with ChatGPT. This new app is the fastest-growing consumer application ever launched since it gained 100 million active users within just 2 months of launching.
The AI chatbot has captivated consumers with its possibilities and yet it seems not many have considered the implications. As a piece of AI technology it learns only by absorbing what is put in. OpenAI – the company behind the phenomenon – said they fed the tool something like 300 billion words from the internet. This included personal information for which they did not get consent. A blog post, an item review, a comment on social media could all very feasibly have been fed to this technology without consent.
What is the problem?
Some may wonder why this could be an issue. One of the main factors is that due to the way that the data was fed OpenAI have no way of knowing whose information is what. So if you were to approach the company and ask what data they hold on you they have no way to sift through the 300 billion words that they used and identify what is yours and what is not. This is a key aspect in data privacy laws, specifically GDPR and there is an ongoing conversation about whether or not the app is GDPR compliant.
The Conversation have also shown that you can very easily get the app to repackage copyrighted information. When they asked for a section of Catch-22 it was repeated to them perfectly.
Another concern that needs to be considered is that anything fed into the application becomes a part of the database. So once information, documents, code, writing, applications have been fed into the database there is no protecting it. Not only could it then be replicated and reused for someone else needing the same thing but it also uses anything in there to continue to learn and improve. Any information you feed it, including personal and sensitive information, becomes part of a database that over 100 million people are actively using.