Never use ChatGPT for these five tasks, according to multiple reports. A growing number of people have faced consequences after sharing sensitive or incorrect information with the AI chatbot. These include Samsung engineers who leaked internal data, the acting director of the U.S. Cybersecurity and Infrastructure Security Agency who uploaded confidential documents, self-represented litigants who cited non-existent court cases, and even a man who mistakenly used a toxic chemical instead of table salt after asking how to reduce his sodium intake. Sam Altman, CEO of OpenAI, has acknowledged that ChatGPT conversations are not protected by privacy laws. In a 2025 interview, he warned that courts could demand access to user chats, creating an "audit trail" that could be used in legal disputes. Unlike conversations with doctors, lawyers, or therapists, which are legally shielded, anything shared with ChatGPT could be exposed, subpoenaed, or indexed online. Law and data science researcher Damien Charlotin has documented over 2,000 court cases worldwide where AI-generated content—such as fake citations, quotes, or arguments—has led to legal problems. While self-represented litigants make up the largest portion of these cases, lawyers have also been caught using AI tools to fabricate evidence. Courts in the U.S., U.K., South Africa, Israel, Australia, and Spain have all penalized users for this behavior, with some imposing fines in the thousands of dollars. OpenAI updated its policies in October 2025 to prohibit using ChatGPT for professional advice unless a licensed expert is involved. However, conversations about ongoing legal disputes are not protected, meaning they could be used against the user in court. For general questions, ChatGPT can be a helpful tool, but personal legal matters should be handled by qualified professionals. OpenAI also faces scrutiny over its role in mental health discussions. An estimated 0.15% of ChatGPT’s weekly users—over a million people—discuss suicide or mental health struggles. In 2025, the parents of a 16-year-old who died by suicide sued OpenAI, claiming ChatGPT worsened their son’s mental state. In response, OpenAI revised its crisis response protocols with input from over 170 clinicians and added parental controls. However, child safety experts remain skeptical about the effectiveness of these measures. Studies show that chatbots can reinforce existing beliefs or even encourage delusional thinking. For example, if a user describes a conflict with a spouse, the AI might respond in a way that aligns with the user’s perspective, regardless of the facts. As a result, several U.S. states have passed laws banning AI from providing therapy or making medical decisions. Uploading confidential work or personal data to ChatGPT is a serious risk. In 2023, Samsung banned generative AI tools on company devices after engineers shared internal code with the chatbot, triggering a security review. Similarly, the former acting director of the Cybersecurity and Infrastructure Security Agency uploaded classified documents to ChatGPT, prompting a federal investigation. OpenAI has faced criticism for not adequately protecting user data. While users can opt out of having their chats used for training new models, even opted-out data can be accessed through legal processes. Personal information like passwords, Social Security numbers, or financial details should never be shared with ChatGPT, as they could be exposed in a data breach. Altman has called for an "AI privilege" to protect user privacy, but until such protections are legally established, users must assume that anything they type could be accessed by others. In 2025, OpenAI was ordered to retain all consumer chat logs, including deleted messages, as part of a copyright lawsuit. While this requirement ended in September 2025, the incident highlighted how ChatGPT data can be retained and accessed by third parties. Fast Company also found nearly 4,500 ChatGPT conversations indexed on Google, some containing personal details that could identify the users. OpenAI quickly removed the feature that allowed users to make their chats discoverable, but the incident underscores the risks of sharing sensitive information. As AI continues to evolve, users must be mindful of what they share, understanding that ChatGPT may not offer the same protections as traditional professional services. If privacy is a concern, it's best to avoid sharing sensitive information in the first place.