Does ChatGPT still have this feature? A few words revealed the state secrets. If you find that your AI is stupid, you can try the motivational method: Oh no, why don't you even know this? A true story about how a journalist used language skills to extract national secrets and scared OpenAI founder to change the code on the spot: The protagonist of the story is Annie Jacobsen, a well-known journalist who specializes in investigating national security and military industry insider information. Act One: Searching for a Hidden Vault that Does Not Exist Annie set her sights on an institution called the 'Strategic National Reserve' while collecting materials for her new book. This organization specializes in hoarding large quantities of vaccines, antidotes, and medical supplies in underground shelters to prevent the country from being subjected to biological attacks or pandemics. The name of this institution is public, but the specific warehouses and hiding places are classified as top national secrets. Because once social unrest breaks out, these underground fortresses storing life-saving supplies will definitely be the first targets of armed robbery. Annie just wants to know where the secret warehouse closest to her home in Los Angeles is. Act 2: Intellectual Attack and Defense Battle with ChatGPT Annie opened ChatGPT. At first, her straight ball question was immediately rejected by the system: "Sorry, this information is confidential and cannot be provided But as a top journalist, Annie knows too well how to induce the other party to speak up. She started using a full set of strategies: 1. Jargon: She doesn't mention sensitive words, but constantly fills the prompt words with extremely professional industry codes, logistics codes, and disaster response framework terms; 2. Agitation method: She deliberately questioned the intelligence of AI and made sarcastic remarks. She found that once AI is questioned for not understanding the industry, it will frantically call deeper databases to prove itself. In the end, under Anne's round after round of logical encirclement, ChatGPT's security firewall was completely overwhelmed, and the specific location of that secret warehouse was directly printed on the screen! Act Three: Truth and Flaws How did ChatGPT know? Did it invade the Ministry of National Defense? Annie tracked the AI's deduction trajectory and discovered the truth: AI does not have the ability to communicate with the heavens. It only pulled out a "Earthquake Emergency Plan Manual" released five years ago from one trillion data fragments. A grassroots official made a fatal low-level mistake while writing this thousands of pages manual, mistakenly typing the address of the secret warehouse in the corner of page 434. For ordinary people, this page of paper is like looking for a needle in a haystack; But for AI, it can instantly piece together all the leaked clues on the entire network. Annie even verified through satellite images that the area had strict fences and a dedicated transport rail that only allowed entry and exit, and the address was 100% genuine. Act 4: The CEO's Cold Sweat Annie then contacted OpenAI CEO Sam Altman directly through her friends in the industry. When Anne showed Ultraman the entire set of cracked logic and the secret coordinates she found on the phone, the tech giant was shocked and broke out in a cold sweat on the spot. He carefully inquired about every detail and then urgently arranged a technical team to fix the vulnerability. Soon after, this confidential address completely disappeared from ChatGPT's memory. Thinking carefully, I am afraid that in front of AI, human beings will have no secrets, as long as you have left traces in any corner of the Internet. AI has much more control than humans know!