AI-“Journalist” News Site Pushes the Interests of OpenAI's Super PAC

Translated by a model; the original is in Russian.

· Artificial intelligence
The outlet Acutus passes AI-generated text off as independent journalism. An investigation ties it to a political machine funded by OpenAI's inner circle.

cover.svg

Investigator Tyler Johnston of the outlet Model Republic has laid out the inner workings of the site Acutus (acutuswire.com) — an anonymous “news” resource that, by all appearances, does not employ a single living journalist.

The trigger was an email received by Nathan Calvin of the advocacy group Encode. A reporter named “Michael Chen” asked for a comment on a story about an AI bill, insisting on a written Q&A format only and disclosing in advance a headline with an obviously predetermined slant. A check through the Pangram detector showed that the email was entirely AI-generated. No person by that name could be found.

What the source code hides

Acutus launched on December 29, 2025, and in less than four months published 94 pieces — on AI policy, Senate races, pharmacy benefit manager reform, crypto regulation and more. Of the 94 articles, Pangram flagged 69% as fully written by AI and another 28% as partially so; only three, by the detector's assessment, were written by humans.

The site's operators left traces right in the open code. The JavaScript revealed an editorial interface with “AI Background Context” and “Question Prompts” fields, along with “Generate Story Draft” and “Regenerate” buttons. The public endpoint /api/wire serves up the entire database along with internal logs: the multi-stage “editorial review” is also carried out by a neural network. According to the timestamps, the whole editing cycle takes 44 seconds on average, with publication happening ten seconds after that. For 42 of the 94 pieces, the automated reviewer itself assigned a “needs revision” status — they were published anyway.

Quotes from “living experts” are obtained in various ways: some are simply scraped from previously published texts, others gathered through that same “Michael Chen” bot. Harvard Business School professor Joseph Fuller, for example, publicly thanked Acutus for the chance to share his opinion, unaware that he had been dealing with a program.

Whose mouthpiece is it

The author links the site to Patrick Hines, head of the Republican PR firm Novus Public Affairs. Acutus's subject matter suspiciously overlaps with the firm's client list: pieces favoring pharmaceutical companies, the crypto lobby, real estate developers and a number of Republican Senate candidates in 2026. Hines himself appears once in an Acutus article as an “independent expert” — with no mention that his firm is, by all appearances, the very outfit behind the outlet.

The thread can be traced further: among Novus's clients is Targeted Victory — a consulting firm whose head became a co-founder of the super PAC Leading The Future. That committee, with a budget of $125 million, is funded by OpenAI president Greg Brockman and the investor a16z, and is overseen by OpenAI's chief political strategist Chris Lehane. Acutus's AI articles echo this committee's talking points verbatim: attacks on Anthropic, criticism of any AI regulation, and accusations that journalists are inciting violence against OpenAI.

The irony is that OpenAI formally prohibits the use of its products for political campaigns, while a warning about the risks of AI-driven influence campaigns recently vanished from its safety framework document. Johnston sums it up: what we have here is exactly such a “Potemkin” newsroom — and it lies outright by calling its texts “expert” and “independent journalism.”


Source: Hacker News

Comments

Sign in, to comment.

  • No comments yet.