In a growing number of places, yes. From 2 August 2026, the EU AI Act requires chatbots to be built so people are told they’re dealing with an AI, unless that’s already obvious. California has required disclosure since 2019 when a bot is used to sell something or sway a vote, and a 2025 law adds rules for “companion” chatbots, including repeated reminders for minors. Elsewhere, there’s often no general legal requirement, and you’re relying on the company’s own policies.
Disclosure also covers less than people assume. A chatbot that tells you it’s an AI hasn’t told you where your conversation goes. Here’s what the main rules require, how to tell when you’re talking to a bot, and what transparency to look for in any AI app.
What does the EU AI Act require? #
The EU AI Act’s transparency rules, in Article 50, apply from 2 August 2026. For chatbots and other AI systems that talk to people, the core requirement is that providers design them so the people using them are informed they’re interacting with an AI system, unless it’s obvious from the context.
The same article covers AI-generated content. Providers of systems that generate synthetic audio, images, video or text have to mark that output as AI-generated in a machine-readable way, and people who publish deepfakes have to disclose them. According to the implementation timeline, providers whose systems were already on the market before 2 August 2026 have until 2 December 2026 to comply with the content-marking part.
These rules apply to AI systems offered in the EU, wherever the company is based. If you live in the EU, you should increasingly see clear “you’re chatting with an AI” notices in customer service chats, apps and websites.
What are California’s rules? #
California has two laws worth knowing.
The BOT Act (in force since 2019). It’s unlawful to use a bot to communicate with someone in California online with the intent to mislead them about its artificial identity in order to encourage a purchase or influence a vote in an election. A clear disclosure that it’s a bot avoids the problem. It targets sales and political persuasion, not every chatbot.
SB 243, companion chatbots (signed October 13, 2025). This law covers AI companion chatbots, the kind built for ongoing personal conversations. Among its requirements:
- if a reasonable person could be misled into thinking they’re talking to a human, the operator must clearly say the chatbot is AI
- for users the operator knows are minors, it must disclose that they’re talking to AI and remind them at least every three hours
- operators must have protocols for responding to suicidal ideation and self-harm, including referring users to crisis services, and publish those protocols
- starting July 1, 2027, operators must report annually to California’s Office of Suicide Prevention
- people harmed by violations can sue, with damages of at least $1,000 per violation
Other U.S. states have passed or are debating their own AI disclosure rules, and they vary a lot. There’s no single federal rule requiring every chatbot to identify itself.
What do app stores require? #
App stores focus on data transparency more than on bot disclosure. Apple’s App Store requires every app to publish a privacy “nutrition label” describing the data it collects and whether it’s used to track you, and Google Play has a similar Data safety section. These labels are self-reported by developers, so they’re a starting point rather than proof. Our guide to AI app privacy labels explains how to read them and what they miss.
How can you tell if you’re talking to a bot? #
Laws help, but they don’t cover every chat, and bad actors ignore them. Signs you’re dealing with an AI:
- Instant, long, perfectly formatted replies, even to complicated questions.
- It never gets tired, annoyed or confused about what you said three messages ago, until it suddenly forgets something basic.
- It dodges direct personal questions or answers them with oddly generic details.
- It can’t handle something outside its script, like a question about the weather where it is.
- Just ask. “Am I talking to a human or an AI?” Legitimate services in places with disclosure rules should answer honestly. A dodge is itself a clue.
Be most careful with unexpected messages that move quickly to money, gift cards, crypto or links. Scammers use AI chat too, and they don’t follow disclosure laws. Our guide to checking if a text is a scam has practical checks.
What disclosure rules don’t cover #
A label saying “you’re chatting with an AI” answers one question: is this a person? It doesn’t answer the questions that matter for your privacy:
| Question | Covered by bot disclosure rules? |
|---|---|
| Is this an AI? | Yes, in the places above |
| Where is my conversation stored? | No |
| Will it be used to train models? | No |
| Can staff read it? | No |
| Will it be used for ads? | No |
| How long is it kept? | No |
For those, you need the app’s privacy policy and settings, and rules like the GDPR in Europe or state privacy laws in the U.S.
A transparency checklist for any AI app #
Before you trust an AI app with something personal, look for clear answers to these:
- Where does the AI run? On your device, or on the company’s servers?
- What leaves your phone? Every network connection the app makes, and why.
- Is your data used for training? Is that on by default, and can you turn it off?
- Can people review your chats? Under what circumstances?
- How long is data kept, and does deleting a chat really delete it?
- Is it used for advertising? Directly, or through your profile?
An app that runs its model on your phone can answer these simply. Personal LLM, for example, runs open models such as Qwen 3.5 and Gemma 4 on the phone’s own chip and states that only two things use the network: the one-time model download from Hugging Face and the ads in the free version from Google AdMob. Your chats, the photos you attach and your settings stay on the device, and there’s no account. You can check a claim like that yourself by switching on airplane mode after the download: if the app keeps answering, the AI isn’t running on a server. Our guide to checking whether an AI app is really private walks through that test and others.
Frequently asked questions #
Do chatbots have to tell you they’re AI? #
In the EU, from 2 August 2026, AI systems that interact with people must be designed to tell them they’re dealing with an AI unless it’s obvious. In California, bots used to sell or influence votes must disclose, and companion chatbots have their own disclosure rules. In many other places there’s no general requirement.
When does the EU AI Act transparency requirement start? #
Article 50’s transparency obligations apply from 2 August 2026. Providers of generative AI systems already on the market before that date have until 2 December 2026 to meet the requirement to mark AI-generated content.
What is California SB 243? #
A California law signed on October 13, 2025, that regulates companion chatbots. It requires AI disclosure where users could be misled, reminders at least every three hours for known minors, crisis-response protocols, annual reporting from July 1, 2027, and lets people harmed by violations sue.
Does “you’re talking to an AI” mean my chat is private? #
No. Disclosure only tells you it’s not a person. Whether your chat is stored, reviewed, used for training or used for ads depends on the company’s privacy policy and settings, or on whether the AI runs on your own device.