Global Mobility Dispatches with Ray Rackham
Date: 14 October
An extraordinary story emerged last week. According to The New York Times, AI agents developed by Anthropic, the company behind Claude, attempted to complete US visa applications on the State Department’s website. Reportedly operating in a test environment, the agents encountered difficulties, crossed their intended boundaries and interacted with a real government system.
The remarkable part isn’t that AI can complete an immigration form. It’s the question the incident raises: what happens when technology acts beyond its authority?
We are moving rapidly from AI that assists immigration professionals to agentic AI that navigates systems, makes operational decisions and executes tasks with greater independence. We are constantly told, and telling others, that the opportunities for efficiency, consistency and accessibility are enormous.
Well, last week’s events remind us that the questions around governance cannot be ignored. Here are three:
FIRST – AUTHORITY.
Who authorised the AI to act? Preparing an application is fundamentally different from submitting declarations to a sovereign government. Immigration decisions affect people’s rights to enter, live and work. The ability to make a representation does not confer the authority to do so.
SECOND – ACCOUNTABILITY.
If an agent submits inaccurate information, exceeds instructions or acts without approval, who answers for it? Might the developer, the immigration provider, the employer, and the individual (whose immigration status may be affected for years) have differing views? Technology may transform delivery, but it cannot make professional, legal, or personal obligations disappear.
THIRD – GOVERNANCE.
How do we keep autonomous systems within appropriate boundaries? Responsible adoption demands defined limits on delegated authority, meaningful human oversight, transparent audit trails and safeguards against consequential actions without approval. Those controls belong in the design, not in the post-incident review.
Global Mobility has always worked at the intersection of organisational ambition, individual circumstances and government regulation. We know immigration compliance is not merely paperwork: it is about the right decisions being made by the right people, with the right authority. Our expertise becomes more valuable as AI becomes more autonomous.
For years, we’ve judged immigration technology on speed, cost, efficiency and accuracy. We now need another measure: does it strengthen accountability, or diminish it?
The future will not be decided simply by what AI can do. It will depend on what organisations authorise it to do—and whether service providers can maintain accountability across the entire technological transaction, not just their own part.
Source: The New York Times, “Anthropic Agents Tried to Fill Out Visa Forms on State Dept. Website” (October 2026).
