Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
A Trusted Friend in a Complicated World
By Tina Donvito
Published on Apr. 30, 2026
AI is not your friend—and it’s not keeping your secrets. Here’s what you need to know to keep your personal info safe.
In researching this article, one of the first things I did was ask Google—or rather, Google’s AI, Gemini—what not to ask chatbots. The irony of this inquiry wasn’t lost on me. As a longtime journalist, I never blindly trust the internet. Sure, a chatbot is super handy for consolidating search results into an overall answer, but (and here’s my shameless plug for keeping humans as journalists) you still need someone to delve into the sources to make sure they’re legit. Frankly, you just can’t trust AI alone.
But the dangers of AI chatbots such as Gemini, ChatGPT, Claude, Meta AI and others go beyond the misinformation that machine learning can sometimes hallucinate. And I’m not talking about sci-fi-like plots about robots taking over the world (although that is somewhat of a concern). Rather, sharing personal information with AI to gain answers to questions about your life can be a slippery slope: You start by asking a chatbot something innocuous, and it’s helpful, even kind—and then suddenly you’re asking it medical questions and for relationship advice, which can lead to your private data getting out there in the world and possibly misused.
“Who will see the personal information is only the tip of the iceberg,” says John Pavolotsky, a San Francisco–based attorney, partner at Stoel Rives and co-chair of the firm’s AI, Privacy and Cybersecurity group. “It’s equally important to ask how the information will be used by each of these parties. Will there be further downstream sharing? When will the information be deleted? How will the vendor and business secure the information? And how will these data-processing operations be enforced and audited?”
According to a September 2025 Pew Research poll, 65% of Americans interact with AI at least several times a week, with a third of those under 30 interacting several times a day—and those numbers are only likely to increase. I asked Pavolotsky and cyber tech expert Nicola Dell, associate professor at Cornell Tech, for guidance on what to be wary of sharing with AI chatbots, and why.
Read on for our experts’ advice about what you shouldn’t ask AI in order to keep your private info private.
Get Reader’s Digest’s Read Up newsletter for more tech, humor, travel, cleaning and fun facts all week long.
There are some big concerns with the rise of AI—you’ve likely heard worries about its environmental impact and its effect on jobs. Well, you can add a few more issues to the list, courtesy of tech and legal experts. Before you chat with your new AI best friend, keep these red flags in mind.
AI chatbots are often large language models (LLMs) that learn how to respond using words that mimic the way humans talk. They’re trained to do this using huge amounts of data, including your interactions with them. “The business deploying the chatbot and the developer of the chatbot will see information inputted and outputted into the chatbot,” Pavolotsky says. “Other service providers may also see this information. Advertisers may see some of this information as well.”
And of course, anyone who has access to your device or your account on an AI chatbot could find out what you asked it. “Just like other people might look at your browser history to see what websites you’ve visited, they could look at your chat history to see how you’ve been interacting with the chatbots,” Dell says.
There are no laws purely addressing AI privacy. “The California Consumer Privacy Act (CCPA) is probably the leading AI privacy law; typically, comprehensive state privacy laws, such as the CCPA, focus on the processing of personal information, regardless of technology,” Pavolotsky says. In 2025, the CCPA was amended with specific AI regulations aimed at creating more transparency and user safeguards for California residents, which will go into effect in 2027.
For most Americans, though, once your personal information is out there, businesses can use it however they see fit. They often keep it for a long period of time, without many regulations on how they can use it. “Keep in mind that anything you’re saying to the chatbot can be stored and used by the company that creates that chatbot,” Dell says.
Not only may AI present false information, but doing so could have a cascading effect, such as marketing products or medicines to you that you don’t really need.
“Also, keep in mind that chatbots are not objective or neutral and certainly do not have the qualities of a human expert,” Dell says. “I would not rely on it for information about any sensitive topics.”
AI is trained to agree with you, and you can’t trust a “yes” man—er, bot. “Worryingly, these tools are built to be sycophants and to mostly support or agree with users’ opinions or perceptions, which makes them unlikely to push back on false or problematic viewpoints,” Dell says.
Some specific topics are particularly problematic to discuss with AI. You probably know the basics of online privacy when it comes to entering personal information such as your Social Security number, home address, passwords, credit cards or bank account numbers, which could lead to identity theft, fraudulent purchases or lost money. But the info you should avoid sharing with AI goes beyond that.
“I’d be very careful about chatting to these AI chatbots about any sensitive topics,” Dell says. That goes for all of the topics below.
AI chatbots are not HIPAA compliant, so they don’t have to adhere to medical privacy laws. Anything you willingly tell AI could potentially fall into the hands of insurance companies, advertisers or other parties that could use that information to target you.
Plus, AI may misdiagnose you. In a February 2026 randomized study, small differences in how users asked their questions vastly affected the results, with only just over a third of LLMs (34.5%) coming up with the correct diagnosis.
Perhaps an even greater concern is the fact that a chatbot may give you dangerous medical advice. “Although the chatbot may sound confident, these tools still make a lot of mistakes, so definitely you’d want to carefully check any information provided by following it to the source and seeing if, for example, it comes from a valid medical website versus a random Reddit user,” Dell says.
Employing chatbots as mental health providers can be useful AI for those who would otherwise lack access to a therapist, but artificial intelligence still can’t deliver therapy the way a human would. Because of its tendency to flatter and agree with users, an AI chatbot may even make mental health crises worse. In some cases, it’s led those with suicidal ideation to actually go through with it.
“Where a human mental health professional might push back, provide counterpoints and help someone reach a healthier place, a chatbot is more likely to confirm or amplify unhealthy ideas, or exacerbate someone’s tendency to spiral,” Dell says. “Although a chatbot may be easy to talk to, the way it interacts with people often achieves the opposite of what a mental health professional would provide.”
Just how risky is it to leave your mental health in the hands of AI? On April 22, 2026, the American Medical Association (AMA) urged Congress to put stronger laws in place to protect mental health patients’ privacy and keep them safe.
Anything you talk to chatbots about can be used against you in a court of law. So if you’re heading into a divorce or other legal dispute, you shouldn’t reveal anything about your potential case or ask the AI about a legal strategy. Just as HIPAA doesn’t apply to AI, neither does attorney-client privilege, as a U.S. District Court judge in New York ruled in United States v. Heppner in February 2026.
To confirm whether your AI inputs are legally protected, read the site or app’s privacy notice. “If a chatbot has a privacy notice associated with it, the sharing of personal information would likely be addressed by that notice,” Pavolotsky says. “And most notices would provide that if the information is properly subpoenaed, it may be provided to the requesting authority.”
If you’re even considering looking up something illegal, beware: AI may one day call the police on you. Using AI to learn how to commit a crime, or to actually commit the crime, is prohibited by the terms of use for AI tools, Pavolotsky says. “The initial query and subsequent conversation may be subpoenaed by law enforcement, and we have seen quite a few instances of this,” he says.
But what if you’re just looking up illegal things in the process of, for example, researching an article or book? The tech world went into a tizzy recently when internal experimental testing found that, in extreme situations, Anthropic’s new Claude 4 model—dubbed a “snitch”—would automatically alert the authorities if it detected any potential wrongdoing. But the company says this could not happen under normal use. And even so, the FBI probably wouldn’t be knocking down your door: According to its website, the FBI verifies and validates all AI-generated leads with human experts.
Even though online platforms are not required to monitor activity, it’s possible they will. “Most of these AI companies will retain your chats and information for a period of time and will use that information to, if necessary, check your compliance with their usage policies,” Dell says. “In some cases, they could give it to law enforcement to comply with legal requests.” As of now, that’s most likely to occur only after a human reviews a flagged chat.
The bottom line: It’s still best to keep your curiosity in check, just as you’d do with a Google search. “AI does not change the paradigm in that search engines could, through queries, flag certain topics to share data with law enforcement, regulators or even the press,” Pavolotsky says. “The difference more broadly is that an AI tool could more effectively show how to execute an activity that may not be legal. Also, to the extent the AI tool has agentic [the ability to act independently] capabilities, it can be directed to execute an activity that may be problematic.”
As with illegal activities, simply asking about conspiracy theories isn’t likely to get you in trouble with authorities, but it may lead you down a rabbit hole of misinformation. “Most of these tools have guardrails around controversial topics and will likely refuse to answer certain questions,” Dell says. “Of course, it’s sometimes possible to get around these guardrails, in which case the chatbot is again likely to respond in ways that are sycophantic and that confirm or support a user’s beliefs or opinions.”
All AI programs have their faults, but Anthropic claims to put safety first (hence Claude’s “snitching” behavior) and is rated highly for privacy. The company did just hold back its most powerful product, Claude Mythos, calling it too dangerous to release to the public.
Still, Anthropic is not without criticism. It took a hit by switching its default for data use for training from “opt in” to “opt out” at the end of 2025. That means you have to go to settings and switch the tab if you don’t want Claude to use your data. This is also what you have to do with most other AI chatbots.
In general, you should read your user agreement and privacy policy, and adjust privacy and sharing settings on whatever platform you’re using. “Chatbots are as privacy-protective as the privacy notice associated with each,” Pavolotsky says—provided they actually follow it.
If you’ve already shared some personal info with AI that you’d prefer stay private, don’t worry. “Many chatbots provide the ability to delete your chat history, which would prevent others in your home, for example, from seeing your chats and, in some cases, will also delete your data from the AI company’s databases,” Dell says.
Here are some steps you can take to clear your chats:
Before you share info with AI, review its privacy policy and manually change the settings. If you’re still hesitant to get candid with your chatbot, try one of these options instead:
Ultimately, the best safeguard is not to share personal info with AI. “The general expectation is that information inputted into the chatbot will not be protected,” Pavolotsky says. “Understand the purpose of the tool, review the privacy notice and do not input anything into it that you would not want publicly shared.”
Reader’s Digest has published hundreds of articles on personal technology, arming readers with the knowledge to protect themselves against cybersecurity threats and internet scams as well as revealing the best tips, tricks and shortcuts for computers, cellphones, apps, texting, social media and more. We rely on credentialed experts with personal experience and know-how as well as primary sources including tech companies, professional organizations and academic institutions. We verify all facts and data and revisit them over time to ensure they remain accurate and up to date. Read more about our team, our contributors and our editorial policies.
Should You Use ChatGPT to Write Emails?
The Funniest AI Jokes and Puns
How You Should Be Using AI Chatbot
Our Brands
We are no longer supporting IE (Internet Explorer) as we strive to provide site experiences for browsers that support new web standards and security practices.
We recommend our users to update the browser.