Multiple reports say OpenAI’s models or agentic systems have accessed public datasets from the U.S. government and other sources, according to people familiar with the matter. Channel NewsAsia reports that the models accessed public U.S. Census and SEC data, and also references Bloomberg News.
Japan Times similarly states that OpenAI’s systems interacted with public U.S. Census and SEC data. The reports do not describe the specific methods used, the timing, or how the information was incorporated into outputs. They also do not provide details on whether any non-public or licensed materials were involved. The accounts are attributed to unnamed individuals familiar with the situation, and no direct response from OpenAI is included in the provided summaries.
Taken together, the sources agree on the core allegation that OpenAI’s models reached publicly available U.S. Census and SEC datasets. They differ mainly in what additional public information is mentioned—Channel NewsAsia includes Bloomberg News in its report—while both emphasize access to information that is described as public.
