Signal President Warns: AI Chatbots Could Be a Privacy Nightmare — Here’s What to Do

Every time you type a question into ChatGPT, Google Gemini, or another AI chatbot, you are handing over more than just a query. You may be sharing personal thoughts, work documents, or even financial details. And according to a recent warning from the president of Signal, the encrypted messaging app, most users have little idea how those conversations are stored, shared, or used.

Signal’s president, who leads one of the most privacy-respecting communication platforms available, has publicly raised concerns about the data practices of popular AI chatbots. The warning, reported by SC Media on June 22, 2026, adds a credible voice to a growing unease among privacy experts.

What Happened

In a brief statement covered by SC Media, the head of Signal cautioned that AI chatbots routinely log and retain user conversations in ways that can compromise privacy. The exact wording of the statement was not available in the original RSS summary, but the core message is clear: even when companies claim to anonymize data, the risk of re‑identification remains real.

Signal itself is built on a foundation of end‑to‑end encryption and minimal data collection. Its president’s critique carries weight because the app’s own architecture shows that strong privacy is technically possible. The contrast between Signal’s design and how most chatbot services handle data is the heart of the concern.

Why It Matters

AI chatbots are not like messaging apps. When you send a message on Signal, only you and the recipient can read it. When you talk to a chatbot, your conversation is typically stored on company servers, analyzed to improve the model, and sometimes even reviewed by human trainers. There is usually no end‑to‑end encryption, and the companies’ privacy policies often allow broad data use.

The risks go beyond theory. Past incidents have shown that anonymized datasets can be matched to real people. If you share a medical symptom, a work project, or a personal problem with a chatbot, that information could reappear in unexpected places — a data breach, a targeted ad, or a future version of the model that remembers things you thought were private.

Signal’s president is not alone in raising alarms. Similar warnings have come from researchers and, in a different context, from military officials debating how AI should be used. But the consumer‑focused warning from a privacy leader like Signal cuts through the noise for everyday users.

What Readers Can Do

You do not have to stop using AI chatbots to protect yourself. A few practical steps can dramatically reduce your exposure:

  • Avoid sharing personal identifiable information (PII). Never input your full name, address, phone number, social security number, financial account details, or passwords. Treat the conversation as though it might be made public.
  • Use temporary or incognito chats. Many chatbot services offer a mode that does not save conversations to your history. Enable it when discussing anything sensitive. Note that “temporary” does not always mean “not logged” — check the privacy policy.
  • Review the privacy policy before using a new tool. Pay attention to sections on data retention, sharing with third parties, and how the company uses inputs for training. If the policy is vague or overly broad, treat the service accordingly.
  • Consider local or offline AI tools. Some AI models can run entirely on your own device (like small language models from Ollama or GPT4All). No data leaves your computer, which eliminates many privacy risks.
  • Keep using encrypted messaging for truly private conversations. When you need to discuss something sensitive with another person, stick with Signal or another end‑to‑end encrypted app. Do not rely on chatbots as a substitute.

Sources

  • SC Media, “Signal president warns about AI chatbot privacy risks | brief | SC Media,” June 22, 2026. (Original report; exact quote not available in RSS summary.)

This article was written based on the published report and publicly known information about Signal’s privacy model. Nothing here should be taken as legal or security advice; see the sources for the original context.