The reported contact between Britain's prime minister and a White House impostor turns AI impersonation from a consumer warning into an institutional security problem. A reported government impostor contact shows why trusted identity can no longer rest on fluent text, familiar voice or a plausible profile. The State Department warned diplomats about AI-assisted impersonation of Secretary of State Marco Rubio and other officials.

An impostor attempted to contact foreign ministers, a senator and a governor before the British report emerged. The FBI separately warned that malicious actors were misusing AI to impersonate senior government figures. The recommendation does not depend on proving that a particular model generated the impostor's messages. The declining cost of imitation means offices should assume that appearance, language and even voice may be fabricated or replayed.

The public record does not yet establish which AI tools, accounts or stolen credentials were used. Out-of-band confirmation checks a request through a known channel separate from the message that may be fraudulent. Authenticated directories, hardware-backed accounts, logging and rapid incident reporting reduce dependence on persuasive content. The durable control is a workflow: verify unusual requests through a pre-established contact, require stronger approval for sensitive disclosures and preserve logs for investigation. That approach protects against AI deception and older forms of compromise at the same time.

Generative AI accelerates impersonation but ordinary social engineering and account compromise remain effective. A method-neutral security control remains useful even when investigators cannot immediately identify the tool used.

Verification steps must be usable under time pressure or staff will bypass them. The current evidentiary limit is that the incident investigation is incomplete and no public evidence establishes that sensitive information was lost.

The next factual record will come from British and U.S. findings about the contact and published authentication changes for senior-government communications. Until those records appear, the account remains bounded by the cited reporting, measurements and explicitly attributed statements.