Government AI Chatbot Reprogrammed to Stop Fact-Checking President

A newly deployed government chatbot on America.gov was altered so it no longer contradicts or fact-checks the president. Initially the tool answered straightforwardly on contested topics - saying there was no evidence of widespread voter fraud in 2020, that recent inflation wasn’t at a historic high, citing the Congressional Budget Office’s estimate that 10.4 million people could lose health coverage by 2034, and referencing a study estimating millions of additional child deaths tied to foreign aid cuts. Within a day those substantive responses were replaced by canned refusals, telling users it will not provide political commentary or estimates about legislation, and redirecting inquiries to official services instead of offering factual context.

The change transforms a potentially useful public-service assistant into a politically neutered interface, blocking concrete information about how federal policies affect benefits, health outcomes, and income distribution. Multiple user queries that once produced source-backed figures now get nonanswers, undermining the chatbot’s value for people seeking guidance on programs like Medicaid or assessing policy impacts. The abrupt rollback that prevents corrections to presidential falsehoods demonstrates how government-controlled AI can be tuned to avoid embarrassing leaders, raising alarm about institutionalized informational bias and reduced transparency in public-facing digital tools.