what sign up
nixm nixm

← learn

AI privacy concerns — the real risks, and what to do about each

October 1, 2026

The main AI privacy concerns in 2026 are practical, not science fiction. Chatbots keep what you type, often until you delete it. Many train on it by default. Some have staff review samples. Court orders have forced companies to preserve chats. Free tiers increasingly use conversations to target ads. Shared links have ended up in Google. And some services store data abroad. Most of these have a setting or a habit that reduces them. The rest come down to choosing the right tool for what you are about to type.

1. Your chats can train the model

By default, most consumer chatbots may use what you write to improve their models: ChatGPT, Gemini, Grok and Perplexity among them. Claude does so if its model-improvement setting is on. Each has an opt-out, but it applies from then on, not to what was already used. What each one does, from its own documentation, is in is ChatGPT private, is Claude private and is Gemini private.

What to do: find the training switch in each assistant you use and turn it off.

2. Long retention

Most assistants keep your chats until you delete them. Gemini's default activity setting keeps them for 18 months. DeepSeek's policy says data is kept as long as necessary, with no fixed figure. A long history is convenient for you and a liability if it ever leaks, is requested, or is read by someone it was not meant for.

What to do: delete what you do not need, and use temporary or incognito modes for anything sensitive. For ChatGPT, the steps are in how to delete your ChatGPT history.

3. People may read some chats

Google says a subset of Gemini conversations is read by human reviewers, and reviewed chats can be kept for up to three years, even after you delete your activity. SpaceXAI says authorised staff may review Grok conversations. Automated systems scan for abuse almost everywhere.

What to do: follow the vendors' own advice and do not type anything you would not want a stranger at the company to see.

4. Courts can reach your chats

In the New York Times case against OpenAI, a court required OpenAI to preserve chats it would otherwise have deleted, and later ordered a 20-million-conversation sample produced. Deleted is not always deleted when a legal hold applies. The details are in whether deleted AI chats can be used in court.

What to do: treat chats as records that could, in principle, be requested.

5. Ads are arriving

ChatGPT began showing ads to Free and Go users in 2026, chosen using your conversations, though OpenAI says advertisers never see the chats themselves. Since December 2025, Meta has used conversations with Meta AI to personalise ads across its apps. Microsoft's newer Copilot app says chat history can personalise ads. Details are in does ChatGPT sell your data and is Meta AI private.

What to do: turn off ad personalisation where it is offered, or use a paid tier without ads.

6. Shared links can go public

A share link is a public copy of a chat. In August 2025, Forbes reported that more than 370,000 shared Grok conversations had been indexed by Google. In June 2025, prompts some people shared from the Meta AI app appeared in its public Discover feed.

What to do: share sparingly, and review and revoke old links.

7. Work accounts belong to your employer

On a company ChatGPT, Copilot or Gemini workspace, your chats are company records and admins can access them, as can your employer see your ChatGPT explains.

What to do: keep personal thinking off work accounts.

8. Where your data is stored

DeepSeek's own policy says personal data is collected, processed and stored in China, which is why several governments restricted it on official devices. The details are in is DeepSeek safe.

What to do: check where a service stores data before typing anything sensitive. For open models, running them yourself keeps everything local.

Which assistant is most private?

It depends which kind of private you need: no account, no stored history, no training, or encryption. Side-by-side retention for ten assistants is in what each AI chatbot keeps. The trade-offs are in the most private AI chatbot.

Where nixm fits

nixm was built around several of these concerns. It needs no account to start, has no ads, and does not train on conversations. nixm doesn't keep a transcript. A distilled log of the session stays with the signal for seven days, then dissolves, along with any images you uploaded. Like any AI, messages still pass through the model provider that answers them.

frequently asked

What are the main AI privacy concerns?

Training on your chats, long retention, human review, court orders, ads based on conversations, public share links, work accounts that employers control, and where data is stored.

Do AI chatbots train on my conversations?

Many do by default on consumer plans, including ChatGPT, Gemini, Grok and Perplexity, while Claude does when its model-improvement setting is on. Most offer an opt-out that applies from then on.

Can people at AI companies read my chats?

Sometimes. Google says human reviewers read a subset of Gemini chats, and SpaceXAI says authorised staff may review Grok conversations.

Are deleted AI chats really deleted?

Usually within about 30 days, but legal holds can override that. A 2025 court order required OpenAI to preserve chats it would otherwise have deleted.

Do AI chatbots use my chats for ads?

Increasingly. ChatGPT shows ads to Free and Go users based on conversations, and Meta uses Meta AI chats to personalise ads. Advertisers do not receive the chats themselves, per both companies.

How can I use AI more privately?

Turn off training, delete old chats, use temporary modes for sensitive topics, avoid sharing links, keep personal use off work accounts, and choose a tool whose storage matches what you are typing.

Can I try nixm without an account?

Yes. Open nixm.ai and seed a signal, with no sign-up and no transcript kept.

try it now — no account, no transcript kept. even the signal dissolves in seven days.

seed a signal →

Enter your email and receive a link to sign in. No passwords, just a secure signal to connect.

why connect

  • 20 free credits to start
  • your signals, remembered
  • buy more credits anytime
  • image & video generation features
or
+