New AI Rules Take Effect: Three Settings to Change on Your Phone Today
Archi4 min read
Contents
The Rule That Changes Your Chat Window#
A transparency rule for general-purpose AI systems took effect this morning in the European Union, and its first visible effect is a label. Chatbots and image generators that serve EU users must now tell people, at the point of first use, that they are interacting with a machine. Providers that skip the notice face fines scaled to global revenue, with the first enforcement letters expected within weeks.
The rule is part of the AI Act's transparency obligations, which apply to systems already on the market. For anyone who uses a chatbot, an image generator, or an AI feature baked into a phone or browser, the practical question is simple: which settings should change today? Several, and most take under five minutes.
Check where the data goes#
Most consumer AI tools offer a setting that keeps your conversations out of model training. Turn it on. Some apps bury it under a privacy menu, others under account settings. If the app does not offer the switch, treat every prompt as public and permanent.
The same applies to AI features inside office suites and email clients. A growing number of tools summarize messages or draft replies by default. Check whether that processing happens on the device or on a server, and whether the vendor retains the text. Retention windows of 30 days are common. Indefinite retention is not rare.
Turn off automatic memory#
Several assistants now remember facts between sessions by default. That is convenient for a recipe and risky for a work document. Open the memory or personalization panel and delete anything you did not deliberately save. Then decide whether new memories should be created automatically.
Audit what the app can reach#
AI features often ask for access to photos, contacts, or your calendar. Grant only what the task needs. A writing assistant does not need your location. An image tool does not need your microphone. On iOS and Android, the permission screen shows every grant; revoke the ones that do not match the feature you actually use.
What the Label Does Not Tell You#
The new notice says a system is AI. It does not say which model, which version, or what data trained it. It does not disclose accuracy rates or known failure modes. Regulators have said the disclosure is a floor, not a full accounting.
That gap matters for anyone who relies on these tools for medical questions, legal forms, or financial planning. A label at the top of a chat window is not a quality signal. It is a reminder to verify.
The phishing angle#
Security teams have flagged a second effect. The notice gives attackers a new pretext. A message claiming to be an official AI transparency notice, asking users to confirm their account to keep access, is already circulating in at least two languages, according to researchers who track phishing kits. Legitimate notices appear inside the app. They do not arrive as links in unexpected emails.
If you receive one, do not click. Open the app directly and check for a banner or a settings page.
Who Has to Comply, and When#
The transparency duties apply to providers placing systems on the EU market, regardless of where the company is based. Smaller providers get a longer runway for some documentation, but the user-facing notice applies now.
Enforcement is handled by national authorities. In practice, that means a patchwork: some countries have staffed their AI offices, others are still hiring. The first fines are likely to target clear, repeated failures rather than edge cases.
What to watch in the next 30 days#
Three signals will show whether the rule bites. First, whether major apps ship the notice as a persistent banner or a one-time popup that users can dismiss and never see again. Second, whether the training opt-out is on by default or hidden. Third, whether app stores begin requiring disclosure in listings.
Users can influence all three by reporting apps that lack the notice. Most national regulators now have an online complaint form. It takes a screenshot and two minutes.
A Short Checklist for Today#
Open every AI app you use. Find the training and memory controls and set them deliberately. Review permissions and revoke anything unrelated to the feature you use. Check whether the app displays the machine-interaction notice; if it does not, note the version and report it. Finally, treat any unexpected email about AI rules as hostile until proven otherwise.
The rules will keep changing. The settings behind them change more slowly, which is why the five minutes you spend now are worth more than the next round of headlines.
More articles