What you should NOT share with ChatGPT

Sensation of the moment, the ChatGPT it's being used in a variety of ways — from entertainment, to study, to work. But we are still talking about a technology that does research and stores information in a database, that is, it can be — and believe me it is — saving information in your interactions. With that in mind, the question remains: what wouldn't it be nice to share with ChatGPT? We will present here some reasons, including those mentioned by the OpenAI, creator of ChatGPT, to prevent sensitive data from being improperly shared and consequently used.

Watch the video on the Showmetech channel:

The best assumption is that anyone in the world can read anything you put on the internet – emails, social media, blogs, LLMs – never post anything you don't want someone else to read.

Gary Smith Fletcher Jones, professor of economics at Pomona College and author of Distrust: Big Data, Data-Torturing, and the Assault on Science.

beware of hackers

Despite being a more general tip, it's still good to talk about it: it's possible for someone to break into ChatGPT's servers and steal your data. There is always that risk of data exposure through bugs and hackers when using an online service, and ChatGPT is not out of that list. In March 2023, for example, a bug was discovered in ChatGPT that exposed the titles, the first message of these new conversations and also payment information from users of Chat GPT Plus.

Nader Henein, vice president of privacy research at Gartner Research, a company that has two decades of experience in enterprise cybersecurity and data protection, says that all of this information you are inputting is highly problematic, as there is a good chance it is susceptible to machine learning attacks.

Chats are always saved

Another piece of information that is very relevant: some OpenAI employees have access to user content. When accessing the page ChatGPT FAQ, we see OpenAI's statement saying that all user content is stored on their systems and the systems of other “trusted service providers”, all of which are located in the United States.

Although OpenAI removes personally identifiable information before obscuring it, the content still exists in its raw form — as it arrived at its source — on these servers. Some authorized OpenAI employees have access to user content for some specific reasons according to the company, one of them being “fine tuning” of your models, unless users prefer not to.

Conversations are used by AI

O ChatGPT it works through chats in which there are conversations between the user and Artificial Intelligence, and these interactions are used for AI machine learning. according to OpenAI usage policy, they make it clear that “they will possibly use information provided by the user to improve their technologies”. But this only works if you let it. There is the possibility of accessing the chat settings and requesting that none of this happens.

On another occasion, users could only choose not to share their data through a Google form that is linked on the FAQ. Now OpenAI has introduced a clearer and easier way to disable data sharing, in the form of a setting within your ChatGPT account. Even with this newanonymous mode“, the conversations are still stored on OpenAI servers for a period of 30 days. And beware, as the company has little information on how they keep your data safe.

We share content with a select group of trusted service providers who help us deliver our services. We share the minimum amount of content necessary to achieve this goal and our service providers are bound by strict confidentiality and security obligations. We do not use or share User Content for marketing or advertising purposes.

OpenAI on sharing content with third-party companies

ChatGPT at work

ChatGPT and several other generative artificial intelligences are being massively used around the world, mainly for those whose work can have some kind of advantage with technology. But not everyone likes the novelty. A Samsung, for example, prohibited its employees from using ChatGPT to check their codes — something quite common in the area — and the practice is subject to a warning.

Other financial companies also got in on the act, such as Citigroup, Bank of America and JPMorgan, as they infringe regulations on third-party messenger apps. Apple also made this type of ban on its employees. This all shows a great concern of large companies with the consequences brought by the ease that this Artificial Intelligence brings. So before using the ChatGPT at work, evaluate whether the content you are inserting there is not sensitive or even ultra sensitive.

ChatGP Therapy

A survey by darkness, a health technology company, revealed that one in four Americans have a certain tendency to talk to an artificial intelligence chatbot instead of going to a therapist. As it is a new and powerful technology, of course there have been cases of users using ChatGPT as a way to nucleoside, or seek some kind of help for substance addiction. These examples show heavy use cases, and how ChatGPT can be a useful, unbiased, and anonymous conversation partner. But it is also worth remembering that these deepest confessions are stored on some server.

As much as technology appears to be quite intelligent, we cannot deceive ourselves, as it still remains just a robot trained to give automatic responses. When we refer to the need to go to a therapy session, we are talking about a patient and a qualified professional to assist him. In addition to the robot not supplying this scenario, your information – again here, sensitive – is being shared and stored.

How to protect yourself in ChatGPT

There are two ways in ChatGPT itself that help you to prevent information shared with Artificial Intelligence from becoming even more vulnerable. The first is to turn off the option “Chat History & Training“. The tool description says: “Save new chats in this browser to your history and allow them to be used to improve our templates. Unsaved chats will be deleted from our systems within 30 days. This setting does not sync across browsers or devices."

By turning this option off, your conversations will still be saved within the 30 day period, but will not be able to be used to train the language model. To deactivate it, click on your name, which is in the lower left corner of the site, go to “Settings", choice "data controls” and disable the first option, “Chat History & Training“. The other alternative is to access the top option, “General", and click on the button "Clear” to delete all existing chats.

General Tips

We've been surfing the internet for years and we know that we should be as careful as possible, especially with what we put on websites and apps. Here's one list of sensitive content and personal information that we should avoid sharing without being sure it's the right and safe place to do so:

• Sensitive personal information;
• Passwords and login data;
• Credit or debit card numbers;
• Bank account details;
• Confidential work or company information;
• Home or work addresses;
• Compromising photos or videos;
• private medical information;
• Trade secrets or business strategies;
• Copyrighted content without proper permission.

And you, what did you think of these tips? Will you be able to use the tool more safely? tell us us Comment!

See also:

ChatGPT App for iPhone it's already available. See how to use it!

Pig iron: Mashable

reviewed by Glaucon Vital in 30 / 5 / 23.