Attention! Secrets that cannot be told to AI

        ChatGPT crashes, did you also crash? In April 2025, the ChatGPT web version suddenly crashed, causing collective anxiety among office workers on social media platforms. A user wailed that the project proposal was stuck halfway through! Some people bluntly say that they can't work today, indicating that generative AI is gradually becoming an indispensable assistant and the most trusted digital colleague. But while enjoying convenience, have you ever thought about what AI knows to keep confidential?

        Some things need to be learned to be 'silent'

                Generative AI learns from a large amount of data to generate responses, and the questions or data input by users may be recorded by the platform and even become materials for training the next generation of models. Although suppliers such as OpenAI and Google have emphasized the security of data processing and provided options such as "not storing chat content" or "deleting conversation records", there have still been cases of information leakage. According to a report by The Wall Street Journal, Jennifer King, an artificial intelligence expert at Stanford University, warns that accidentally entering sensitive data while interacting with AI tools may pose privacy risks, and special attention should be paid to the following five types of information:

        1. Personal Identifiable Information (PII)

        Personal information such as ID number, passport number, date of birth, residential address, phone number, etc. can be directly identified. Once it falls into the hands of malicious individuals, it may be stolen or subjected to social engineering attacks.

        2. Medical and health information

        Many people will ask AI to read health examination reports, explain physical symptoms, and even analyze the diagnostic content of doctors. However, currently most AI tools on the market have not been included in medical regulations and lack sufficient protection mechanisms for health data. If you wish to request AI assistance in interpreting the inspection report, it is necessary to cover sensitive fields such as name, ID number, and medical institution, and only retain necessary data content to reduce the risk of personal data leakage.

        3. Financial account and transaction information

        Account number, bank code, credit card information, investment details, tax information, salary records, etc. are all contents that should not be entered. Once these materials are leaked, it is not only a matter of security, but also likely to cause financial losses.

        4. Internal company information

        Many people provide the project name, customer information, and internal process content to AI for assistance in writing emails, reports, and simplified reports. Seemingly convenient, but very dangerous. Some AI platforms have explicitly revealed that the usage content of free or open versions may be used to improve models.

        5. Login password and authentication credentials

        More and more AI tools can help write programs, string applications, and even simulate task automation processes. Some users, in order to save time, may directly provide API keys, passwords, and tokens to AI, causing extremely high risks. Once credentials are recorded, leaked, or misused, it can lead to account theft or system intrusion.

        Five tips to make AI use with peace of mind

        1. Avoid using authentic information

        Please replace real information with fictional content or simulated scenarios. If you need to present the actual situation, be sure to cover up sensitive information such as name, account number, and amount in advance to avoid personal data leakage.

        2. Enable incognito chat mode

        ChatGPT has a circle in the upper right corner of the user interface, which can be checked to activate the "Temporary Chat" function, similar to a seamless window, to prevent conversation content from being recorded or used for model training.

        3. Make good use of privacy settings function

        Most generative AI tools provide privacy control options, such as whether to store chat content, whether to use it for model training, etc. It is recommended to first go to "Account Settings" to select more rigorous privacy protection functions.

        4. Regularly clear usage records

        Regardless of which platform you use, you should develop the habit of deleting conversation records to avoid leaving too many traceable digital traces, in order to reduce the risk of data misuse or leakage.

        5. Choose to use the enterprise's self built platform

        Far Eastern New Century Corporation has imported AIFE 3.0 (Love Far Eastern Group AI Platform) internally, which maintains the efficiency of AI collaboration while providing a trusted usage environment, making AI tools truly digital partners in the workplace.

        Clever use of AI, don't let input become output

        Generative AI has become an important force in the modern workplace, covering almost everything from communication to creation. But don't forget, its' intelligence 'comes from every piece of data we input. The more you give, the stronger the effectiveness; The finer the information, the higher the risk. Making good use of AI to make work more efficient is the key to investing time in more valuable areas. After all, even the smartest tools require users to handle them with wisdom. Choosing the right tools, using the right methods, blocking risks, and keeping efficiency to oneself are the most robust ways to enter the AI era.

        *Image source: freepik

        #