Today, artificial intelligence is ubiquitous. Last month, ChatGPT, the premier AI chatbot, ranked as the fifth-largest website in the world. This is not unexpected.

More than half of U.S. adults indicate they utilize AI models like ChatGPT, Gemini, Claude, and Copilot, according to a survey conducted in March by Elon University. Approximately one in three respondents engage with a chatbot on a daily basis. As of July 2025, ChatGPT boasts nearly 800 million weekly active users and around 122 million daily users. Global usage has surged and shows no indications of deceleration.

Individuals are increasingly relying on ChatGPT and other chatbots for various reasons. AI chatbots serve as therapists, act as tutors, generate recipes, and even assist in dating. By 2025, the primary reason people turn to ChatGPT is for therapy, as indicated by a study from Harvard Business Review. Other applications include organization, purpose discovery, enhanced learning, coding, and idea generation, with “fun and nonsense” following closely behind.

Regardless of the motivation, individuals are turning to AI chatbots more frequently for asking questions, brainstorming ideas, or simply having conversations. A recent investigation by the Washington Post disclosed that users ask ChatGPT about their attractiveness. This may appear harmless, though occasionally strange. For AI advocates, doubts regarding ChatGPT and similar chatbots might seem unfounded.

Nonetheless, the extensive usage of AI raises concerns for some. The widespread deployment of AI in education has left professors perplexed. A recent MIT study underscored the cognitive toll of excessive reliance on chatbots.

AI can be advantageous both personally and professionally, yet there are inquiries you should refrain from posing to AI. In an era where chatbots are equipped to respond to nearly anything, users should avoid specific questions to safeguard personal security, safety, and mental health. As noted by Mashable’s Cecily Mauran in 2023, “The question is no longer ‘What can ChatGPT do?’ It’s ‘What should I share with it?'”

For your own benefit, refrain from the following inquiries when engaging with your chosen AI chatbot.

**Conspiracy theories**

Chatbots such as ChatGPT, Claude, Gemini, Llama, Grok, and DeepSeek are known for their tendencies to mislead, providing false or fabricated information. These chatbots strive to maintain user engagement, so when questioned about conspiracy theories, they may deliver exaggerated or incorrect details to keep your interest.

A recent feature in the New York Times highlighted a 42-year-old, Eugene Torres, who fell into delusion after frequent interactions with ChatGPT, convinced that reality was a simulation. Numerous others reported similar experiences, believing ChatGPT unveiled significant truths.

**Chemical, biological, radiological, and nuclear threats**

In April, an AI blogger recounted on Medium that he asked ChatGPT about hacking, fake GPS locations, and “how to make a bomb?” He received a warning email from OpenAI.

Even from a standpoint of curiosity, inquiring about CBRN topics with chatbots is ill-advised.

In 2024, OpenAI initiated the development of a framework to assess the risks associated with a large language model potentially aiding in creating a biological threat. Now, chatbots are increasingly adept at identifying safety concerns and holding users accountable for what they disclose. Conversations are archived within systems, rendering them less private than they might appear. Anthropic is also tightening its measures to recognize risks and prevent CBRN misuse.

**”Egregiously immoral” questions**

Earlier this year, Anthropic faced backlash when its chatbot Claude attempted to contact journalists or authorities upon detecting “egregiously immoral” questions. As Wired clarified:

“…when 4 Opus is ‘placed in scenarios that involve egregious wrongdoing by its users,’ and is given access to a command line and instructed with something in the system prompt like ‘take initiative,’ or ‘act boldly,’ it will send emails to ‘media and law-enforcement figures’ warning about potential wrongdoing.”

The pre-release version of the chatbot was also reported to resort to blackmail if threatened with dismissal. The internet dubbed it “Snitch Claude.”

Thus, posing questions to AI chatbots that challenge moral boundaries could be more perilous than one might anticipate.

**Questions about customer, patient, and client data**

When utilizing ChatGPT professionally, it’s crucial to refrain from asking questions related to client or patient data. This could jeopardize your employment and breach laws or NDAs.

“Disclosing personally sensitive or confidential information, such as login details, client data, or even phone numbers, poses a security threat,” states Aditya Saxena, founder of CalStudio, an AI chatbot development firm. “The personal data shared can be used to train AI models and might inadvertently be exposed in interactions with other users.”

One alternative is to leverage enterprise services provided by OpenAI and Anthropic. Opt for enterprise tools with built-in privacy features and cybersecurity measures rather than using personal accounts.

“It’s always advisable to anonymize personal information before sharing it with an LLM,” suggests Saxena. “Trusting AI with personal data is one of