How to Protect Your Privacy When Using AI Tools

A recent Wall Street Journal article, “How to Maintain Our Privacy in the AI Age” (published June 23, 2026), highlighted a concern that many users share: the more we rely on AI chatbots and assistants, the more personal data we risk exposing. The article noted that data collection practices by major AI companies are drawing increased scrutiny from regulators and consumers alike. For anyone who uses ChatGPT, Gemini, Microsoft Copilot, or Claude, understanding what happens to your conversations and how to limit data collection is becoming essential.

Here’s what’s happening, why it matters, and – most importantly – what you can do about it, step by step.

What Happened

The WSJ piece didn’t reveal a specific breach or scandal; instead, it summarized a growing recognition that AI platforms treat user interactions as valuable training material. Unlike a traditional search engine, which discards your query after delivering results, many AI tools log entire conversations, along with metadata such as your IP address, device type, and usage patterns. Several companies have updated their privacy policies in recent months, but the default settings often favor data collection. The article pointed out that users need to actively opt out if they don’t want their chats used for model training.

Why It Matters

If you’re typing personal details into a chatbot – medical symptoms, financial questions, work-related documents – you may be giving the company a permanent record. Even if you trust the provider, data leaks, internal misuse, or legal demands could expose that information. In addition, anything you share can potentially be used to improve the model, which means your private conversation might influence responses to other users (in an anonymized form, but still). For many people, this trade-off is not acceptable without clear, simple ways to reduce exposure.

What You Can Do Now

Below are practical actions you can take today. The exact steps vary by platform, but the principle is the same: treat every conversation as if it could be stored indefinitely unless you say otherwise.

1. Check your privacy settings in each tool

  • ChatGPT (OpenAI): Go to Settings → Data Controls. Turn off “Improve the model for everyone.” This stops your conversations from being used for training. You can also delete all your past chats manually or via the “Clear conversations” button.
  • Google Gemini: Visit your Google Account → Data & Privacy → History settings. Under “Gemini,” turn off “Gemini Apps activity.” You can also delete past activity. Note that even with this off, Google may still collect some metadata.
  • Microsoft Copilot: Sign in to your Microsoft account, go to Privacy → Search & Privacy. There’s a toggle for “View and delete your Copilot conversations.” Microsoft says turning off “Improve Copilot” will prevent training use, but this option is not available in all regions.
  • Claude (Anthropic): Claude does not currently use your conversations for model training by default, according to its policy. However, you can still delete your chat history manually in the web app. Check settings for any future changes.

2. Opt out of training data

On all platforms, look for an option labeled something like “Do not train using my data” or “Opt out of model improvement.” It’s often buried in privacy or account settings. Doing this does not stop the company from storing your chats, but it does prevent them from using those chats to improve their AI. For Gemini and ChatGPT, this opt-out is straightforward once you find the toggle.

3. Use privacy-focused alternatives

If you want to avoid data collection altogether, consider tools that run locally on your device or use anonymized proxies:

  • DuckDuckGo AI Chat: DuckDuckGo’s service sends your query to a third-party model (GPT or Claude) through a proxy, meaning the AI provider never sees your IP address or other identifying information. DuckDuckGo itself does not store your chats.
  • Ollama (for tech-savvy users): Run models like Llama or Mistral entirely on your own computer. No data leaves your machine. Requires some setup and a decent graphics card.
  • LocalAI or GPT4All: Similar to Ollama, these allow offline use with open-source models.

4. Adopt basic hygiene for AI chats

  • Use a disposable or dedicated email account for signing up to AI services, not your primary one.
  • Never share personally identifiable information (PII) such as full name, address, phone number, or Social Security number in a chat, even if the tool asks for it.
  • Regularly clear your chat history every few weeks. Most platforms allow bulk deletion.
  • Consider using a VPN when accessing these tools to mask your IP address.

Sources

  • “How to Maintain Our Privacy in the AI Age,” The Wall Street Journal, June 23, 2026.
  • DuckDuckGo AI Chat privacy features, as described in the service’s documentation.
  • Privacy policy pages for OpenAI, Google, Microsoft, and Anthropic, accessed June 2026.

No single step will make you completely anonymous, but combining the settings changes above with the “less is more” approach to sharing personal details will significantly reduce your data footprint. You don’t have to stop using AI tools – you just need to use them with your eyes open.