One in Three Users Share Secrets with AI Chatbots That Aren't Private, DuckDuckGo Survey Reveals

AI Chatbots Become Digital Confidants as Privacy Risks Mount

Americans are treating AI chatbots as trusted confidants, sharing personal information they wouldn't reveal to friends, family, or medical professionals. A DuckDuckGo survey of nearly 2,000 U.S. adults found that 32% of users have confided secrets to AI chatbots3

3

. Among self-described AI enthusiasts, that figure jumps to 56%, while 43% of parents reveal things to AI they don't share with the most trusted people in their lives, including their children1

1

.

People turn to ChatGPT, Claude, and other AI chatbots for help with health issues, financial advice, parenting questions, and relationship problems. The conversational nature of these tools creates an atmosphere of privacy that encourages frank disclosure4

4

. Parents show particularly high engagement, confiding in AI almost twice as much as non-parents3

3

.

Most Users Don't Know Chatbot Conversations Are Not Private

The DuckDuckGo survey exposed a critical knowledge gap about privacy risks. Only 47% of respondents understood that most AI chatbots use conversations to train their models by default, while 53% either didn't know or were unsure3

3

. Just 25% knew that AI chats can be subpoenaed by law enforcement1

1

.

Once informed about how their data is used, 58% of people expressed discomfort with their conversations being used to train AI models, with 30% specifying they felt "very uncomfortable"3

3

. The survey revealed that nearly 40% of respondents had no trust in companies to protect their data, while nearly 50% had no trust in the U.S. government1

1

.

ChatGPT Conversations Used as Evidence in Court Cases

A Washington Post review of public records found a dozen instances where chatbot transcripts were cited in court cases over the past two years2

2

. Chatbot logs have been used as evidence in legal proceedings ranging from murder investigations to civil litigation. ChatGPT transcripts appeared in a South Carolina double murder investigation and the Palisades arson case, while Snapchat AI conversations were used in a 2024 murder trial1

1

.

In one civil case, a teenage boy's intimate ChatGPT conversations about a settlement and personal struggles were swept into public court records during litigation against Meta, Snapchat, TikTok, and YouTube4

4

. "A fifteen-year-old may type things into a chatbot that he may not say to a therapist, a parent, or a friend, and he does it without realizing that it could end up in a defense expert's report," said Mike Morgan, a partner at Morgan & Morgan representing the teen4

4

.

Law Enforcement Can Access AI Chatbot Data Through Subpoenas

Companies operating large language models like OpenAI, Anthropic, and Google are legally required to turn over chat data when subpoenaed by law enforcement or the U.S. government1

1

. OpenAI CEO Sam Altman acknowledged this reality on a podcast: "If you go talk to ChatGPT about your most sensitive stuff and then there's like a lawsuit or whatever, we could be required to produce that"1

1

.

In May, the FBI arrested a Florida man after OpenAI reported him for repeatedly telling ChatGPT about plans to rape and murder his ex-girlfriend4

4

. The user, Darren Zhou, had allegedly spent months describing violent plans in detail. He pleaded guilty and was sentenced to eight years of probation4

4

.

OpenAI uses automated detection systems to identify potentially concerning activity and notifies law enforcement when "conversations indicate an imminent and credible risk of harm to others"2

2

. Anthropic similarly discloses information when it has a "good-faith belief" that disclosure prevents serious harm2

2

.

How Companies Use Your Data to Train AI Models

Unless users opt out, AI systems record and save conversations for future use, including training large language models1

1

. When logged into an account, ChatGPT and Claude retain chat history until users delete it. Both OpenAI and Anthropic say they delete conversations from internal servers within 30 days after deletion in most cases2

2

.

Data memorization poses significant privacy risks with large language models. "When someone asks the same question as someone else, it is possible that the LLM gives them exactly the same answer it gave someone else," explained Uttara Ananthakrishnan, assistant professor at the University of Washington's Foster School of Business. "The LLM might not know that the answer contains information that someone can use to identify someone else"1

1

.

Last year, hundreds of chats on Anthropic's Claude appeared in Google search results, demonstrating how sensitive information can be exposed1

1

.

Employers Can Access Enterprise Accounts and Monitor AI Use

Many workplaces have enterprise accounts with OpenAI and Anthropic that prevent AI companies from training models on employee inputs2

2

. However, corporate ChatGPT or Claude accounts are managed by employers, who can access chat history just like email or Slack. Anthropic offers tools for administrators to view chats and files2

2

.

A survey showed that two in five workers share sensitive information with AI chatbots they wouldn't tell friends or family, even putting company information in private chats on personal accounts1

1

. Some major companies including Walmart, Delta, T-Mobile, Chevron, and Starbucks are using AI systems to gauge worker positivity or toxicity1

1

.

How to Protect Your Data When Using AI Chatbots

Both OpenAI and Anthropic offer temporary or incognito chat modes that automatically delete conversations within 30 days2

2

. However, exceptions exist—OpenAI may retain deleted data if an account has been banned for policy violations or if legally required2

2

.

"Unless you are having a chat with a service that has a temporary chat or, basically, an incognito version ... [and] you're also having it within a browser that's not tracking you, the answer is no. You can't be sure that it'll be totally private," said Jen King, a privacy researcher at the Stanford Institute for Human-Centered Artificial Intelligence2

2

.

DuckDuckGo emphasizes that users can still change their approach to privacy settings. "The most valuable information you could share with Big Tech is your next question, not the questions you asked days or months or years ago. So, it's never too late to make the switch to more private services online,"3

3

the company stated. Legal experts recommend choosing services with transparent data-handling policies and avoiding sharing sensitive information through workplace accounts or personal chatbots that could face legal implications of AI in legal proceedings2

2

.