Signal President Warns: How AI Chatbots Could Be Spilling Your Secrets
If you’ve asked a chatbot for help drafting an email, summarizing a document, or even just brainstorming ideas, you’ve probably assumed your input stays private. That assumption may be wrong.
Signal’s president recently issued a public warning about the privacy risks of AI chatbots, drawing attention to how these tools handle—and potentially expose—personal data. For anyone who values privacy, it’s worth understanding exactly what’s at stake and what you can do about it.
What happened
In a statement covered by SC Media, Signal’s president highlighted concerns about how AI chatbot providers collect, store, and share user conversations. While the exact wording of the warning is not fully detailed in the reporting, the core message is clear: many popular chatbots do not treat user input as confidential.
Signal, which has built its reputation on end-to-end encrypted messaging, is in a unique position to comment on data handling practices. The company’s president pointed out that users often treat chatbots like private assistants, unaware that those conversations may be logged, reviewed by human trainers, or used to improve future models.
Why it matters
The risks fall into a few categories:
Data retention and training. Most chatbots rely on large language models that are trained on vast amounts of text, including user conversations. Even if a company says it anonymizes data, there is no guarantee that sensitive information—like medical details, financial data, or confidential work documents—cannot be reconstructed or inadvertently exposed in future responses. OpenAI, for example, has acknowledged that ChatGPT conversations may be used for training unless users opt out.
Third-party access. Many chatbot platforms share data with third-party service providers (for moderation, analytics, or infrastructure). The contracts governing these arrangements are rarely transparent. A user who pastes a confidential email into a chatbot may effectively be handing it to several companies at once.
No encryption guarantee. Unlike a Signal message, which is encrypted end‑to‑end by default, chatbot conversations are typically stored in plain text on the provider’s servers. If those servers are breached, your conversations could become public.
Persistent logs. Some chatbots keep conversation histories indefinitely, creating a permanent record of everything you’ve typed. Deleting a chat from your account does not always remove it from the provider’s backup systems.
Signal’s president is not the only voice raising the alarm, but his position as a leading privacy advocate gives the warning added weight. When a company known for refusing to hand over user data to governments warns about AI chatbots, it’s worth listening.
What readers can do
You do not have to stop using chatbots, but you can reduce your exposure. Here are practical steps that privacy‑conscious users can take today.
1. Check the provider’s data policy. Before using any chatbot, locate its privacy page and look for wording about “training data,” “conversation logs,” and “third‑party sharing.” If the language is vague, treat the tool as unsafe for sensitive information.
2. Opt out of training where possible. Many major chatbot providers offer an opt‑out setting in their account or privacy controls. For ChatGPT, you can turn off “Improve the model for everyone” in your settings. Bard (Gemini) and other tools have similar options, though the exact location changes frequently.
3. Avoid entering personal or sensitive data. A simple rule: do not type anything into a chatbot that you would not write on a public postcard. That includes your full name, address, phone number, passwords, work‑related secrets, and health information.
4. Use temporary or incognito chats if available. Some platforms offer ephemeral conversations that are not saved to your history. Bing Chat (now Copilot) and ChatGPT have modes that stop saving messages, though providers may still retain them for safety checks.
5. Clear your chat history regularly. Even if you cannot opt out of training, deleting old conversations reduces the amount of data tied to your account. Most chatbots have a one‑click delete option in their settings.
6. Consider using local, open‑source models. Tools like Llama, Mistral, or GPT4All can run on your own computer with no data leaving your device. They are less convenient and sometimes less capable, but they offer much stronger privacy guarantees.
7. Be skeptical of “privacy‑friendly” claims. A chatbot that advertises privacy may still log metadata (timestamps, IP addresses, etc.) or rely on analytics services. Look for independent audits or open‑source transparency before trusting a new service.
Sources
The primary source for this article is the SC Media brief titled “Signal president warns about AI chatbot privacy risks,” published June 22, 2026. The specific details about data handling practices are based on widely reported information from OpenAI, Google, and Microsoft about their chatbot data policies as of mid‑2026. Readers are encouraged to review the original SC Media article and each chatbot provider’s privacy policy for the most current terms.