Artificial intelligence regulation is no longer focused only on how systems are built. It now reaches the companies offering them, the people deploying them, the content they create, and the political risks they may bring into elections.
The European Union’s AI Act sits at the center of that shift. Adopted in 2024, it uses a risk-based legal framework, assigning obligations to AI providers, deployers, importers, distributors, and other actors according to the system or model involved. The rules also reach companies outside Europe if they offer AI tools to people in Europe.
What the EU AI Act Requires
One requirement applies to any company offering AI tools in Europe, wherever that company is based: it must mark AI-generated content. The rule connects the use of AI with the need for people to recognize when content comes from a machine.
Anthropic applies these content markings worldwide because it has no reliable way to apply them by region. That choice shows how a European requirement can shape a company’s practices beyond Europe, even when the law itself focuses on AI tools offered in the European market.
The AI Act is part of a wider European approach to online safety and digital responsibility. In August 2026, the EU placed ChatGPT under the strictest tier of its Digital Services Act, Europe’s online safety rulebook. OpenAI, the creator of ChatGPT, must assess and reduce the system’s “systemic risks.”
Those risks include elections and the spread of disinformation. The focus matters because people already use chatbots for political questions: 1 in 5 voters have asked AI chatbots about elections and politics.
Why Political Advice Has Become a Test
The Dutch Data Protection Authority tested four chatbots before the 2025 national election. It found that the political middle ground “virtually disappears” in their voting advice. That finding puts attention on more than false content. It also raises questions about how AI systems present political choices and whether their answers reflect a balanced range of views.
Jordi Calvet-Bademunt, a researcher at Vanderbilt University, is among the entities connected to the discussion of AI and politics. The concern sits within a wider political divide in the United States: a 2025 Pew Research Center survey found that 8 in 10 Americans believe Republican and Democratic voters cannot agree on basic facts.
That gap gives political AI rules a difficult job. Regulation must address election risks and deceptive content, but different governments have chosen different methods. In 2024, Thierry Breton, the EU’s top digital rules enforcer, reminded Elon Musk of his legal obligations before Musk’s live interview with Donald Trump on X.
The exchange showed how digital rules can apply to political events hosted on online platforms. The EU AI Act and the Digital Services Act do not stand alone in this debate. They sit alongside decisions about platform responsibility, election information, and the boundaries of political speech.
Deepfakes, Government AI, and Free Speech
In the United States, dozens of states, including California and Hawaii, have passed laws restricting AI-generated political deepfakes. These laws target deceptive AI content about candidates, but courts have blocked laws in California and Hawaii that ban sharing such content about candidates, citing the First Amendment.
That conflict captures the challenge of regulating political AI. Governments want to limit deceptive material during elections, while courts have treated some restrictions as a free-speech problem. The result is a policy landscape that differs across jurisdictions.
President Trump issued an executive order in 2025 titled “Preventing Woke AI in the Federal Government.” The order requires the AI models the government buys to be “truth-seeking,” ideologically neutral, and not favoring ideas such as “diversity, equity and inclusion.”
The order places government purchasing at the center of the debate over how AI should answer questions. In Europe, the focus includes risk levels, content markings, and systemic risks. In the United States, the policies described here also address ideological neutrality, government procurement, deepfakes, and constitutional limits.
The worldwide policy picture includes harsher consequences as well. In December 2025, two activists in Vietnam were sentenced to 17 years each in prison. That fact sits alongside the legal and political disputes in Europe and the United States, showing how different governments respond to digital and political activity.
What These Rules Mean for AI Users
For users, the main change is that AI regulation now follows the full path of an AI system. It can cover the company that provides a model, the organization that deploys it, the content it produces, and the risks that emerge when people use it in politics.
The EU framework does not treat every AI system in the same way. Its obligations depend on the system or model involved, which places risk at the center of the legal structure. ChatGPT’s placement under the strictest tier of the Digital Services Act shows how those concerns can grow when a tool reaches political life and public debate.
Election content will remain a hard test. Marking AI-generated material can help people identify it, but political advice, disinformation, deepfakes, and disagreements over free speech raise separate questions. The rules now being developed across Europe, the United States, and other countries will shape how those questions are answered.
Based on
