5 things users shouldn't share when using ChatGPT
Every day, ChatGPT processes approximately 2.5 billion requests globally. Because its responses resemble a friendly conversation, many people inadvertently forget that this is an AI system and are sharing information that shouldn't be disclosed.
Unlike Google or traditional search engines, ChatGPT gives the feeling of talking to a real person. This "closeness" makes users easily lower their guard, sharing personal, financial, or work information without hesitation.
However, AI chatbots are not static systems. They are constantly being improved, learning from input data, and still face certain security risks. Meta AI previously encountered vulnerabilities that allowed access to private messages; ChatGPT also faced issues with prompt insertion and the risk of publicly shared conversations being indexed by search engines.
This means users need to be vigilant when using ChatGPT, even more cautiously than with other online platforms.
1. Personal Information
This is the most common and dangerous mistake. A study by the international cybersecurity organization Safety Detectives showed that in 1,000 public ChatGPT conversations, users frequently shared their real names, addresses, ID numbers, phone numbers, email addresses, usernames, and passwords.
ChatGPT can be very helpful in writing CVs, cover letters, or administrative forms without requiring any real personal data. Using fake information or substitute symbols is both effective and helps avoid the risk of data leakage or public indexing.
Users can also disable the option that allows ChatGPT to use conversations to train models, but this doesn't make sharing personal information safe. The best approach is still to avoid sharing personal information in the first place.
2. Financial Information
Many people view ChatGPT as a “personal financial advisor,” from budgeting to investment analysis. While useful at a general level, OpenAI also acknowledges that ChatGPT can make mistakes.
More importantly, there is no reason to enter bank account numbers, credit card information, investment details, or tax records into ChatGPT. AI systems are not designed to protect financial data like banks or financial institutions.
In a worst-case scenario, this data could be misused for fraud, identity theft, financial fraud, or cyberattacks. If analysis is necessary, use fictitious data or data with sensitive information redacted.
3. Personal medical information
ChatGPT is increasingly being used to find health information. A recent survey showed that 1 in 6 adults use AI chatbots to search for medical knowledge, and this number rises to 1 in 4 among young people.
The issue isn't about understanding general medical concepts, but about sharing personal details such as diagnoses, test results, medical history, or mental health issues. When combined with identifying information, this is an extremely sensitive group of data.
Once entered into ChatGPT, this medical data is outside of health information protection mechanisms and it is virtually impossible to control how it is processed afterward. ChatGPT can help understand medical terminology, but it should never replace a doctor or official medical records.
4. Confidential documents and information
Internal emails, client documents, unpublished reports, project data, or intellectual property should not be included in ChatGPT in their raw form.
Using AI to edit emails or summarize documents might seem convenient, but it could inadvertently expose data to secure processes within a business. This risk is particularly serious in fields such as healthcare, finance, law, and research.
The safety principle is to only import content that has been anonymized, abridged, or is no longer confidential.
5. Anything related to illegal activity.
OpenAI is obligated to comply with requests from law enforcement agencies in the U.S. and many other countries. Sharing, inquiring about, or attempting to circumvent legal restrictions may result in the account being flagged, restricted, or providing data as required by law.
ChatGPT's safeguards against illegal activities, fraud, harm, or malicious code aren't always perfect, and that further highlights the need for users to take responsibility for what they share.
In summary, ChatGPT is a powerful and useful tool, but it's not absolute privacy. The core principle is simple: don't share anything you're not willing to make public or share with third parties.
Using AI chatbots in general, and ChatGPT in particular, intelligently involves not only leveraging their features but also ensuring the security of users' personal data, work information, and digital identities.
