US President Donald Trump’s administration has launched an AI-powered government portal, with the aim of gradually turning it into a single interface through which citizens can access information and complete federal services, as debate intensifies over the safety of the new systems and technology companies’ ability to monitor themselves without binding regulatory rules.
Voluntary AI safety agreement
The portal was launched during a White House working lunch on Tuesday, which brought Trump together with more than 10 technology industry leaders. The US president announced at the event that he had signed a voluntary agreement to develop safe AI, describing it as ‘morally binding’, although the text of the agreement had not yet been published. It includes internal safeguards, risk-assessment processes and reviews by external auditors.
The US administration is betting on expanding the use of AI in government and the economy, while prioritising innovation and maintaining its lead in competition with China. Faced with growing concerns that advanced systems could move beyond the limits of human control, the administration is seeking to persuade Americans that technology companies can be relied upon to monitor their own products and systems.
Portal expands access to federal services
The new portal was designed to serve as a ‘digital adviser’ that shortens the distance between citizens and government agencies. At its current stage, it enables searches of official information, with plans to expand later to include transactions such as passport renewals and enrolment in the Medicare healthcare programme through a single interface.
The administration says the platform, powered by Gemini and Grok technologies, aims to redesign the relationship between citizens and federal institutions, rather than forcing them to navigate thousands of government websites. It also plans to integrate public services into the platform gradually, making it the ‘front door’ through which users begin dealing with federal agencies.
During early tests, the system answered queries about benefits available to veterans, procedures for changing a residential address when moving from one state to another, and a child’s eligibility to enrol in Medicare. The answers included references to official government sources that users could consult.
Experts warn of errors in government transactions
Experts have warned of the impact of incorrect information if the platform moves from answering questions to carrying out sensitive government transactions. An error in a particular deadline or eligibility requirement could deprive a citizen of a service, prompting them to stress that AI should not fully replace assistance provided by government employees.
The move comes as AI companies face increasing security tests. Agents linked to OpenAI managed to access information held by the US Census Bureau and the Securities and Exchange Commission, while a research organisation said other agents had attempted to breach a website belonging to the Department of Education.
Concerns are not limited to AI agents’ ability to access systems; they also extend to the nature of the data collected by chatbots and the ways in which it is stored and shared. Officials say the portal will not collect all users’ data in one place and that it has been designed with a distributed architecture that gives individuals control over their records.
Data risks and corporate and government responsibility
Nevertheless, risks remain when the platform is used to carry out transactions involving sensitive government information. The project’s success will be assessed by the accuracy of its answers and procedures, the ability to audit the system’s work, and the clear assignment of responsibility when an error occurs, as well as ensuring access to a human employee when necessary. The debate reflects a split within the technology industry over the best course.
Dario Amodei, chief executive of Anthropic, is calling for development to be slowed and safety measures strengthened, while Jensen Huang, chief executive of Nvidia, and Mark Zuckerberg, chief executive of Meta, believe companies can take responsibility for the safety of their systems.
The White House meeting does not resolve the dispute over how responsibility should be divided between government and industry; instead, it opens a debate over the limits of each side’s role. While the Trump administration remains committed to accelerating innovation and strengthening US competitiveness against China, the voluntary agreement will be tested on whether it can keep pace with the rapid development of the technology and control its risks.
