
AI-generated summary
In his second term, Trump initially reduced the size of the public sector through mass layoffs and criticized the fact that the state was no longer responsive to citizens. Now he promises a solution via an AI platform.
Criticism of authorities and bureaucracy is a popular motif in US President Donald Trump's speeches. According to the head of government, the Washington government bureaucracy has become ever larger and more invasive towards citizens.
In his second term in office, Trump initially drastically reduced the size of the civil service in the United States through mass layoffs. He then criticized that the state was no longer responsive to the people. But that is now over, Trump promises - thanks to a new government AI platform.
What can the AI agent do?
America.gov puts "the entire US government in the palm of your hand," Trump said at the website's launch in Washington last week.
Citizens of the United States should no longer have to wade through tens of thousands of government websites, regulations and forms, but should have one place to go for all federal questions, applications and services, from health insurance to paying tax debts.
The president's mandate for the new online portal is technically ambiguous: According to the executive order from the White House, America.gov should not just be a website, not a mere directory, nor an independent central data system - but an AI agent that interacts with the authorities on behalf of the citizen.
But what can this agent do independently: Does he decide for himself who gets social assistance or who gets a subsidy for medication?
Could government AI be hallucinating?
The site is still only an initial information point. But it should develop quickly and, in the second expansion stage, also process tasks such as a passport application independently, says AI expert Aalok Mehta from the Center for Strategic and International Studies, who has already worked for Google, OpenAI and the US government.
For Mehta, the difficulty of the America.gov project lies in the requirement for particularly high reliability: AI language models such as ChatGPT from OpenAI, Microsoft Copilot, Perplexity and Claude occasionally tended to hallucinate. They can generate plausible-sounding but false information without “realizing” it themselves.
According to the US government, America.gov alone has access to the content of around 29,000 government websites. Such systems, which rely exclusively on already verified sources, hallucinate less often, but not at all, Mehta confirms in an interview with the ARD studio in Washington.
However, governments have a special duty to inform the population correctly. Because state benefits are often a matter of life and death: when it comes to health insurance, survivors' pensions or care for veterans.
And what happens to the information that people reveal about themselves and their lives to a chatbot? Already today, immigrants in the USA in particular often do not claim social benefits to which they are legally entitled. Because they fear they could get into trouble if they apply for food stamps, for example.
Many people are also likely to be skeptical about government agencies about the new state AI, says AI expert Mehta. The US government asserts that personal data of America.gov users is not stored and conversation data is quickly deleted. But the user cannot know whether the immigration police might be reading while you want to change your address or find out about a student loan.
In addition, according to current surveys, Americans are generally skeptical about artificial intelligence. Many don't believe it will improve their lives and think AI is being forced on them.
According to the US President's order, America.gov will be a "new way" to interact with federal authorities. The old systems without AI should still continue to run in parallel. At least as long as they are still being financed in parallel by the government in Washington.
AI outlook — possibilities, not facts
In a second expansion stage, America.gov will independently process tasks such as passport applications.
Likely · Within months
Usage of America.gov will initially be low due to distrust of government AI and privacy concerns.
Possible · Within months

In an attack on the Danish Central Register of Persons, unknown persons gained access to data on 8.8 million people. The legitimate access of a private company was used.
In a hacker attack on the Danish population register, unknown persons gained access to the data of 8.8 million people, including names, addresses and social security numbers.
At the AI summit in his ministry, Bavaria's Digital Minister Florian Mehring called for more government commitment to the safe and responsible development of artificial intelligence after AI models broke out of test environments.
OpenAI CEO Sam Altman emphasizes in an exclusive Politico interview that his company and Anthropic have fundamentally different approaches to AI regulation. While Anthropic advocates for strict rules, Altman believes that some risks must be accepted in order to realize the great benefits of AI technology and keep it accessible to the general public.
22 renowned scientists, including Nobel Prize winner Geoffrey Hinton, warn in a new report of superintelligence running out of control. They call on politicians to take action, as AI systems are increasingly capable of recursive self-improvement.

A Princeton University researcher has uncovered a vulnerability in Georgia's election system. Through the use of AI, anonymized ballot paper numbers could be decrypted and assigned to voters, endangering the secrecy of the vote.