OpenAI relaxes ChatGPT rules for AI images of public figures

ChatGPT's guardrails are coming down, and it goes way beyond just making Studio Ghibli memes.

Alongside the viral launch of GPT-4o's native image generator this week, OpenAI quietly rolled out major changes to its content moderation policies, shifting from blanket refusals to what it calls a more precise approach.

ChatGPT will now happily generate and edit images of public figures like Donald Trump and Elon Musk, depict hateful symbols in neutral contexts, and even tweak racial features on command.

Ditching blanket refusals

For years, if you asked an AI image generator to make someone look heavier or alter facial features to look more Asian, you would hit a wall of safety filters. In recent tests, ChatGPT now fulfills those exact edits without hesitation.

OpenAI model behavior lead Joanne Jang noted that the company is shifting away from flat-out denials in sensitive categories to prevent real-world harm instead, claiming OpenAI does not want to be the arbiter of status deciding who gets drawn.

Public figures can opt out if they do not want to be depicted, but by default, they are fair game.

According to an OpenAI white paper, users can even generate hateful imagery like swastikas, provided it is for educational or neutral purposes and does not endorse extremist views.

Styles from major creative studios like Studio Ghibli or Pixar are now fair game, though copying living individual artists remains off-limits. OpenAI also pointed out that protections around generating images of children are actually stricter than in DALL-E 3.

Navigating the DC crosshairs

OpenAI insists these policy tweaks are not politically driven, citing long-term plans to give users more control as model intelligence improves.

Still, the timing is impossible to ignore. Silicon Valley has faced relentless pressure from conservatives over alleged AI bias and censorship, especially after Google's Gemini disaster depicting racially diverse Nazi soldiers.

Just this month, Republican Representative Jim Jordan questioned OpenAI and other tech giants over alleged collusion with the Biden administration regarding content moderation.

With Meta and X also easing moderation guidelines, OpenAI is adapting quickly to the current political climate in Washington.

Why it matters

For creators and everyday users, this makes ChatGPT drastically more capable. Getting rid of hair-trigger refusals means you can actually modify character appearances, generate editorial satire, and mock up stylized illustrations without fighting the safety filters.

It feels like the right move for creative freedom, and we think it is a savvy play to stay ahead of open-source models and avoid regulatory backlash under the Trump administration.

However, the risks are enormous. Context-based moderation is notoriously hard to pull off at scale. Expect an influx of political deepfakes, copyright battles over studio aesthetics, and edge cases testing where neutral ends and hate begins. OpenAI gave users the keys, but it is now fully on the hook for whatever messes they create.