How to Use AI Chatbots Without Sacrificing Your Privacy: Tips After Signal’s Warning

In June 2026, Signal’s president made headlines by publicly warning about the privacy risks tied to AI chatbot use. The exact details of his statement are still being parsed, but the core message was clear: the same data‑collection habits that have long plagued social media and messaging apps are now embedded in the tools millions of people talk to every day.

If you use ChatGPT, Gemini, Claude, or any similar service, you are handing over more than just your queries. Every prompt, every uploaded file, and every conversational thread can be stored, analyzed, and potentially accessed by third parties. Below is a breakdown of what’s happening behind the scenes — and, more importantly, what you can do about it.

What Happened

On June 22, 2026, SC Media reported that Signal’s president had issued a stark caution regarding AI chatbot privacy. While the full text of his remarks is not yet public, the warning underscores a growing concern among security professionals: consumers often assume their chats are private when, in practice, they are not.

Signal itself does not operate an AI chatbot. Its encrypted messaging platform is designed with privacy as a foundation. The warning therefore came from a position of authority about the broader ecosystem — reminding users that the same standards do not apply to most AI chatbot providers.

Why It Matters

The privacy pitfalls of popular AI chatbots are not hypothetical. Here is what typical usage exposes:

  • Data logging. Most chatbots retain your conversation history for model improvement and safety monitoring. That includes personally identifiable information, work‑related content, or even sensitive health details you might innocently type.
  • Third‑party access. Many providers share anonymized or aggregated data with partners, researchers, or advertising networks. “Anonymized” does not always mean unrecoverable, especially when enough context is present.
  • Employee review. Some companies allow human reviewers to read snippets of conversations to train or audit the AI. Although often done under nondisclosure agreements, the privacy boundary is thinner than most users realize.
  • No end‑to‑end encryption. Unlike Signal, WhatsApp, or iMessage, most chatbot platforms do not encrypt your chats from your device to the service’s servers — let alone in a way that prevents the provider from reading them.

For everyday consumers, the risk is not that a stranger will peek at your grocery list. It is that a pattern of shared information — workplace data, family updates, passwords typed by mistake — can be stored indefinitely and used in ways you did not consent to.

What Readers Can Do

You do not have to stop using AI chatbots. But you can take concrete steps to limit exposure.

Avoid sharing anything you would not put in a public post.
Treat every chatbot conversation as if it could be printed on a bulletin board. Never include passwords, Social Security numbers, bank details, or private medical information. If you need the AI to analyze a document, redact personal identifiers first.

Turn off chat history and training features.
Most major providers offer a setting to disable conversation logging or to opt out of using your data for model training. On ChatGPT, for example, you can find this under Settings > Data Controls. On Gemini, look for “Activity & history” in your Google Account. These settings are not always obvious, but they are worth the few minutes it takes to locate.

Use pseudonyms and generic references.
When describing a scenario, change names and avoid giving your real location, job title, or other identifying clues. The AI does not need to know your exact situation to provide useful advice.

Consider privacy‑focused alternatives.
Several services now offer chatbot access with stronger privacy defaults:

  • DuckDuckGo AI Chat (free, anonymized, no training on your chats)
  • Brave Leo (built into the Brave browser, with a local processing option)
  • Private GPT‑style models that run entirely on your own computer (tools like llama.cpp, GPT4All, or Ollama)

None of these are perfect, but they reduce the data trail you leave behind. If you need the convenience of a mainstream chatbot, at least use the privacy settings and a browser‑based session that you clear regularly.

Check the privacy policy before using a new tool.
Most people skip this step. Look for answers to:

  • Does the provider retain your conversations? For how long?
  • Is your data used to train future models?
  • Can employees or contractors view your chats?
  • Is the service covered by a data‑processing agreement (relevant for work use)?

If the policy is vague or buried, consider that a red flag.

Sources

  • “Signal president warns about AI chatbot privacy risks | brief,” SC Media, June 22, 2026. Link

  • General privacy research from nonprofit organizations such as the Electronic Frontier Foundation (EFF) and consumer reports from Mozilla Foundation’s “Privacy Not Included” guides.