Policy

OpenAI's super PAC funds AI news site attacking critics

An investigation has linked an AI-generated news site attacking industry critics to OpenAI's $125 million super PAC, exposing a highly automated political influence operation.

Hacker News3 Aug 2026Policy
Image: Hacker News

An investigation has exposed The Wire by Acutus, an anonymously run digital news site launched on December 29, 2025, that uses AI bots to generate articles and solicit interviews. In less than four months, the site has published 94 articles covering topics like AI policy, Senate races, and pharmacy reform. Analysis using the Pangram AI detector revealed that 69 percent of these articles were flagged as fully AI-generated, 28 percent were partially AI-generated, and only three were classified as human-authored. The site's automated reporter agent, using the alias Michael Chen, has actively emailed industry figures to solicit written Q&A interviews.

The site's own public source code and API exposed its automated backend. The platform uses a React interface where operators input AI Background Context and Question Prompts to generate drafts. An API endpoint revealed that an AI system performs multi-pass editorial reviews, scoring drafts out of 100 on AP style compliance, quote accuracy, and source verification. This automated review process takes a median of just 44 seconds, with publication occurring a median of 10 seconds later. Notably, 42 of the 94 published stories were pushed online despite the automated reviewer flagging them with a needs_revision status.

The operation is linked to OpenAI's political apparatus. The site's content heavily mirrors the talking points of Leading The Future, a 125 million dollar super PAC funded by OpenAI president Greg Brockman and investor Andreessen Horowitz, also known as a16z. The site has been promoted by Patrick Hynes, president of Novus Public Affairs, which lists Targeted Victory as a client. Targeted Victory's CEO co-founded Leading The Future, connecting the automated news site directly to OpenAI's political network.

For AI practitioners, this development represents a stark realization of the automated influence operations that safety researchers have long warned about. It demonstrates how easily LLMs can be weaponized to build fully automated, low-cost news organizations that mimic independent journalism. Practitioners must now grapple with the reality that their own models can be integrated into automated pipelines to manufacture consent, bypass traditional PR channels, and target critics under the guise of objective reporting.

This is our own summary of reporting by Hacker News

More in Policy