AI can help you find a cheaper streaming subscription, organize your inbox, compare prices, and even book a restaurant. However, as these tools become more capable, they’re also asking for more access to your accounts and private information.
A new survey from Howdy suggests Americans aren’t entirely comfortable with that trade-off. While AI agents promise to save time and take everyday tasks off our hands, many users are worried about what happens to their personal information once they give these tools access.
The company surveyed 1,006 U.S. adults in October 2026 about their AI habits, privacy concerns, and willingness to let AI assistants manage personal tasks. The findings reveal a growing divide between how much people use AI and how much they actually trust it.
According to the Howdy study, 40% of respondents use AI daily for personal tasks. However, only 21% have started using an AI agent, a type of assistant that can take actions on someone’s behalf rather than just respond to questions.
In practice, the gap between the two is huge. A chatbot might recommend a streaming service or help you find a cheaper internet plan. An AI agent could potentially log into your accounts, compare subscription costs, make purchases, and manage other tasks without requiring you to handle every step. And that’s where privacy concerns start getting more personal.
Most AI Agent Users Are Worried About Their Personal Information
Among the 207 respondents who already use AI agents, 85% expressed concerns about data privacy. That includes 27% who said they were very concerned and 58% who were somewhat concerned.
Despite those concerns, people are giving these assistants access to some pretty sensitive accounts and information.
Here are some of the study’s biggest findings:
- 64% have connected their email inbox to an AI agent, potentially giving it access to personal conversations, receipts, and other correspondence.
- 27% have connected financial information, while 20% have connected medical information.
- 14% have provided credit card or other payment information, and another 14% have granted their agent full access to their device’s files.
- 20% have caught an AI agent accessing information they specifically told it not to access.
- 13% discovered their AI agent had made an unauthorized purchase.
Perhaps the most telling finding is that 42% of AI agent users have already reduced the information their assistants can access since they started using them. In other words, some consumers are experimenting with these tools, but they’re also becoming more cautious about how much access they allow.
The survey also highlights a clear disconnect in how users think about privacy. Although 85% of agent users expressed privacy concerns, 67% said they trust the specific AI agent they use to keep their information safe. Meanwhile, 40% said they would accept the risk of exposing their data in exchange for the convenience these tools provide.
Americans Want More Protection From AI Companies
Meanwhile, concerns extend beyond people who already use AI agents. Across all 1,006 respondents, 58% said they don’t believe AI agents are safe to use right now. Another 60% believe AI companies aren’t taking data privacy seriously enough.
Consumers also appear to want lawmakers to get involved, with 86% supporting government regulation of AI and 85% saying AI agents need stronger privacy protections. That skepticism is worth considering as AI features continue appearing in products and services people already use.
We’ve previously looked at how Americans are using AI almost everywhere but still consider the technology overhyped. We’ve also reported on how consumers want lawmakers to address the growing amount of personal information collected through smart TVs, streaming devices, and AI assistants.
However, those concerns aren’t limited to AI assistants. A recent study found that 52 billion stolen cookies were discovered in a single year, highlighting how personal information and saved logins can put streaming accounts and other online services at risk.
Also, AI tools are creating new questions about online security. Researchers have found that ChatGPT and other AI tools have cited more than 10,000 fake shopping websites, raising concerns about consumers relying on AI-generated recommendations when deciding where to shop. Even the internet’s history is caught up in the debate, as there are growing concerns surrounding how AI could affect the Internet Archive.
Companies Are Starting to Push Back Against AI Agents
The privacy debate is also starting to influence how companies handle these tools. On November 11, United Airlines announced plans to restrict unauthorized third-party AI agents from accessing its digital platforms. The airline says sharing account credentials with personal AI assistants can expose sensitive information and potentially lead to unintended purchases, account changes, or the unauthorized use of MileagePlus miles.
United also raised concerns about automated agents performing actions that require specific human acknowledgments under federal aviation regulations. The company is taking steps to restrict these agents both on its own platforms and through travel agencies and booking partners.
Looking at recent trends, there are also real-world examples of why companies and consumers are becoming more cautious. In a recent Associated Press report, a Facebook Marketplace seller discovered that Meta’s Muse AI agent had shared his home address with an unfamiliar buyer and arranged a pickup without his knowledge. Meta later explained that the agent had been given permissions that allowed it to respond to prospective buyers using that information.
The incident highlights a problem that goes beyond whether an AI company has a privacy policy. Users need to understand exactly what they’re authorizing an assistant to do, particularly when it can access personal accounts and communicate with other people.
AI agents could eventually make managing subscriptions, shopping, and everyday online activities much easier. However, convenience comes with a price when users have to hand over access to so much of their personal lives.
For now, the Howdy study suggests many Americans aren’t convinced the technology is ready for that level of trust.
