How to Keep Your Privacy in the Age of AI: A Practical Guide

If you’ve used ChatGPT, Google Gemini, or Microsoft Copilot recently, you’ve likely typed something you wouldn’t want posted on a billboard. Names, company strategies, personal worries—these tools encourage us to be conversational, but that openness comes with a privacy cost. Nearly every major AI service stores your prompts and uses them to refine its models unless you take specific steps to stop it. Recent reporting, including coverage in The Wall Street Journal, has highlighted how quickly user data can leak or be repurposed in ways people didn’t expect. This isn’t about abandoning AI—it’s about using it with your eyes open.

What Happened

The Wall Street Journal’s article “How to Maintain Our Privacy in the AI Age” (published July 2026) examined the data practices behind popular generative AI tools. The piece noted that even when companies promise not to share your information with third parties, they often retain your conversations for model training by default. Similar investigations have surfaced instances where sensitive business data entered public training sets or where chat histories were exposed through security lapses. Meanwhile, third‑party plugins and browser extensions that promise AI enhancements often have broad permissions to read content across websites, compounding the risk.

Why It Matters

The convenience of AI assistants is real. They save time, help brainstorm, and automate routine writing. But every prompt you feed into a cloud‑based model is data a company can store, analyze, and—depending on its policies—use to improve its product. If that data contains health details, financial information, or confidential work material, the exposure could be serious. Even if a company has robust security, the sheer volume of conversations makes it a tempting target for breaches. For everyday consumers, the main concern isn’t necessarily malicious misuse but the slow erosion of control over personal information. You might not mind an AI knowing your grocery list, but you probably don’t want it retaining your child’s school schedule or your salary negotiation notes.

What Readers Can Do

The good news is that you don’t have to stop using AI. A few deliberate changes can shrink your privacy footprint significantly.

Adjust settings in each tool.

  • ChatGPT (OpenAI): Go to Settings → Data Controls and turn off “Improve the model for everyone.” This stops your conversations from being used for training. You can also delete your chat history manually or set automatic deletion after a certain period.
  • Google Gemini: Under your Google Account → Data & Privacy → “Gemini Apps activity,” you can pause the saving of your conversations. Google also allows you to delete individual entries or clear all history.
  • Microsoft Copilot (and Bing Chat): In your Microsoft account privacy dashboard, you can turn off “Improve AI models” and delete your search and chat history.

Anonymize your prompts.
Avoid typing names, phone numbers, addresses, or any personally identifiable information. Treat every prompt as if it might be reviewed by a stranger. Use placeholders—for example, “My client in the healthcare industry” instead of “Dr. Smith at Northside Clinic.” This habit alone eliminates a large class of risks.

Use privacy‑focused alternatives.
Local models such as Llama (run on your own machine via tools like Ollama or LM Studio) never send your data anywhere. Encrypted services like DuckDuckGo AI Chat act as an anonymizing layer, letting you use ChatGPT or Claude without revealing your IP or identity. They don’t store conversations. These options may be slightly less capable than the full cloud versions, but for everyday tasks they’re often sufficient.

Watch third‑party integrations.
Browser extensions that claim to supercharge AI chat often read the content of every page you visit. Before installing any, check the permissions. If it requires “read and change all your data on websites,” ask yourself whether you trust it with your banking portal and email. A safer approach is to use the tool’s official web interface and avoid extra plugins.

Search wisely.
AI‑powered search in Google, Bing, or Brave can also collect your queries. Use a search engine that doesn’t track you (e.g., DuckDuckGo) for anything personal. If you do use AI search, clear your history regularly.

No single step guarantees total privacy, but these practices together create a meaningful barrier between your data and the models that train on it.

Sources

  • The Wall Street Journal, “How to Maintain Our Privacy in the AI Age,” July 2026.
  • OpenAI privacy policy and settings documentation.
  • Google AI privacy controls (support.google.com).
  • Microsoft privacy dashboard.
  • DuckDuckGo AI Chat privacy overview.