admin
Erica Knight, a conservative communications pro, posted screenshots this week that stopped being amusing fast. Her exchange with OpenAI’s ChatGPT shows the model declining to “strengthen” a promotional post for Vice President J.D. Vance, then happily rewriting a glowing post about Governor Gavin Newsom. The screenshots are real, the concern is real, and Big Tech’s silence so far is not reassuring.
What happened: a screenshot that set off alarms
Knight’s posts on social media show ChatGPT telling her it can’t “strengthen a political endorsement or assess suitability for president” when asked to help polish copy for Vice President J.D. Vance. When given a similar prompt for Governor Gavin Newsom, the model rewrote the post into a full‑throated endorsement. Knight’s caption: “The model favors Newsom and censors JD.” Conservatives are rightly sharing the screenshots and asking why an AI would behave like a polite PR intern who only roots for one side.
Policy vs. proof: OpenAI’s rules, prompt sensitivity, and why one screenshot isn’t the whole story
Let’s be clear: OpenAI publicly bans using ChatGPT to generate campaign messaging. So a refusal to draft an endorsement is consistent with company policy. But policy enforcement is supposed to be neutral, not selective. The problem is the implementation — slight differences in wording, the model version, or hidden moderation signals can make the same tool say “no thanks” to one user and “here you go” to another. One screenshot is a red flag, not a verdict. We need exact prompts, timestamps, and the model version to know whether this was bias, a bug, or just sloppy policy enforcement.
Why conservatives should care about AI censorship and election influence
This isn’t just about hurt feelings or PR copy. When AI systems begin to nudge political messaging, even subtly, they can shape narratives and voter impressions at scale. If a widely used chatbot declines to help one candidate while boosting another, that asymmetry can matter — especially closer to an election. Big Tech already decides which posts trend, which ads run, and which accounts get throttled. Now they host the tools that write our political messages. Conservatives should demand better transparency before machine‑filtered politics becomes the new normal.
What we must demand — transparency, audits, and real accountability
Start with the basics. Ask Erica Knight to provide the exact prompts, timestamps, and the ChatGPT version used. Ask OpenAI to reproduce the interaction, release logs, and explain which safety or political filters were triggered. Push for independent audits of political refusals and for Congress to hold hearings on AI political‑content enforcement. If the tech giants won’t show their work, users should migrate to alternatives and lawmakers should step in. We can tolerate sensible safety rules — but not secret rules that tilt the playing field.
Short of that, conservatives should treat every convenient refusal as suspicious and every convenient endorsement as suspect. AI can be a useful tool — or it can be a referee that favors one team. Right now, we have too many unanswered questions and not enough answers. Demand them, loudly and clearly.
