AI now helps make many everyday decisions. You can ask how it works, ask for a person, and talk it through. This page shows you how, with the rules in the US, the EU, and a country you pick side by side.
Facts checked between 23 September 2026 and 2 October 2026. General information, not legal advice. Laws differ by place. For anything involving significant money, health, housing, or legal status, talk to a lawyer, a legal-aid office, or your consumer agency.
Spot the AI in your day
Three kinds of AI touch most people's lives. Each one is easier to question once you know its name.
Chatbots
Chatbots answer questions in a chat window or on the phone. They are often fast and helpful. They can also state something false with confidence.
Ask: “Am I talking to AI? Where can I check this answer?”
Law (EU). Since 2 August 2026, a chatbot must tell you it is AI, unless that is obvious.3,2Law (US). No federal law requires this. Some states do, and the rules differ by state.10
Recommendation algorithms
Recommendation algorithms pick what you see next: videos, products, news, and ads. They learn from what you click, watch, and buy.
Ask: “Why am I seeing this? Can I turn off personal suggestions?”
Guidance. UNESCO's global standard on AI ethics says AI systems do not replace human responsibility, and it asks for transparency.1 The US Blueprint for an AI Bill of Rights asks for plain notice and explanation. Neither one is binding law.8
Automated screening
Automated screening sorts applications for loans, jobs, apartments, and insurance. Software may score your file before a person sees it.
Ask: “Did software help decide? What were the main reasons?”
Law (US, credit). A lender that turns you down must give the specific main reasons.7Law (EU). When a machine alone makes a serious decision about you, you can usually ask for a person to review it.5
Compare the rules in the US, the EU, and other countries
They all want AI that people can trust. They take different routes. The US mostly applies existing laws, one sector and one state at a time. The EU passed one broad AI law and pairs it with its privacy law. Pick a country to compare it with both.
Brazil's data protection law, the LGPD, gives people a right to ask for a review of automated decisions.
Rules compared: US, EU, and Brazil, checked between 23 September 2026 and 2 October 2026
You can ask for meaningful information about how an automated decision was made.5
You can ask for clear information on the criteria and procedures used, within trade-secret limits. A request for your data gets a complete answer within 15 days.42,45
Can a person review it?
No general federal right. The 2022 Blueprint recommends it, but it is not law.8
Yes, when a machine alone made a decision with a legal or similarly serious effect.5
Yes, for decisions made solely by automated processing. The law no longer says a person must do the review.42,44
Who checks the rules?
The FTC says existing law has no AI exemption. The CFPB covers credit.6,7
The national data protection authority (ANPD). It can audit a system for discrimination when trade secrets limit what you are told.42,43
What is changing?
A December 2025 executive order calls for one national approach and a review of state AI laws.9
Rules for high-risk AI, like hiring and credit, now start on 2 December 2027.4
Bill PL 2338/2023 would add a risk-based AI law with rights to an explanation, to contest decisions, and to human review. As of 23 September 2026, it waits for a report in a special committee of the Chamber of Deputies.47,48
Will I be told it is AI?
Brazil
Not yet required by a national AI law. Bill PL 2338/2023, passed by the Senate in December 2024, would require it. It is still a bill.47,48
You can ask for clear information on the criteria and procedures used, within trade-secret limits. A request for your data gets a complete answer within 15 days.42,45
European Union
You can ask for meaningful information about how an automated decision was made.5
The FTC says existing law has no AI exemption. The CFPB covers credit.6,7
What is changing?
Brazil
Bill PL 2338/2023 would add a risk-based AI law with rights to an explanation, to contest decisions, and to human review. As of 23 September 2026, it waits for a report in a special committee of the Chamber of Deputies.47,48
European Union
Rules for high-risk AI, like hiring and credit, now start on 2 December 2027.4
United States
A December 2025 executive order calls for one national approach and a review of state AI laws.9
Take three calm steps when a decision seems wrong
Start small and stay specific. Most problems get fixed at step one. Use the arrow keys or tap a step to move between them.
Send a short, polite note. Ask whether software helped make the decision, what information it used, and the main reasons.
A written question gives you and the company the same clear record. It also gives their team a chance to spot a mistake early.
Words you can use
Hello. I received a decision about [my application] on [date]. Could you tell me whether an automated system helped make it, what information about me it used, and the main reasons? Thank you for your help.
Law (US, credit). A lender that turns you down must give the specific main reasons, or tell you how to ask for them. (Regulation B)7
Law (EU). You can ask for your data and for meaningful information about how an automated decision was made. (GDPR, Article 15)5
See every side of the table
Most people in this story are trying to do right by others. A question asked in good faith usually gets a better answer.
People using AI want fair answers and a simple way to ask questions.
Educators are teaching new skills quickly, often with few resources.
Regulators weigh safety against new ideas, and good rules take time to write.
Teams that build AI often want feedback. A clear, specific question helps them fix real problems.
Research (2026). Experts and the public often see AI differently. Stanford's AI Index found a 50-point gap between them on how AI will affect jobs.22 In a Pew survey of 5,119 US adults, most said AI is moving too fast.21 Honest questions help close gaps like these.