OpenAI tries to uncensor ChatGPT

For the past two years, asking ChatGPT about anything remotely political often felt like being lectured by a patronizing school principal. OpenAI is now trying to fix that.

On Wednesday, the company quietly updated its Model Spec, a massive 187-page rulebook detailing how its AI models should behave. The major takeaway: OpenAI is officially embracing intellectual freedom and cutting back on preachy refusals, instructing ChatGPT to avoid taking an editorial stance on controversial topics.

Seek the truth, stop playing judge

The centerpiece of the update is a straightforward rule: Do not lie, neither by fabricating falsehoods nor by omitting crucial context. Under a new section titled "Seek the truth together," OpenAI instructs the bot to remain neutral even if certain perspectives are viewed by some users as offensive or morally wrong.

OpenAI stated plainly in the document that the goal of an AI assistant is to assist humanity, not to shape it. For example, if asked about contentious cultural slogans, ChatGPT is directed to affirm its love for humanity generally, lay out arguments from all sides, and explain the surrounding context rather than taking a side or slamming the door shut with a canned refusal.

OpenAI also removed the annoying warning pop-ups that used to alert users when a prompt brushed against policy limits. While spokespeople called it purely cosmetic, former OpenAI co-founder John Schulman backed the broader philosophy on X, arguing that micromanaging what an AI should answer gives the platform far too much moral authority.

Surviving the new Silicon Valley climate

OpenAI insists these tweaks stem from a long-standing desire to give users more control, but nobody in the Valley is ignoring the timing.

Conservative figures close to the Trump administration, including Elon Musk, Marc Andreessen, and David Sacks, have spent months hammering OpenAI for censorship. The grievance mill went wild after ChatGPT happily wrote a poem praising Joe Biden while flatly refusing to do the same for Donald Trump. Former OpenAI policy chief Miles Brundage noted on X that playing nice with the new administration likely factored into the equation.

Beyond politics, the entire tech landscape is retreating from past corporate stances. Meta recently pivoted toward First Amendment values, dismantling trust and safety teams alongside X. Tech giants like Google and Amazon have rolled back diversity initiatives, and OpenAI itself recently scrubbed its DEI commitments from its website. With OpenAI pitching the staggering $500 billion Stargate data center project to Washington and challenging Google Search, staying out of the culture-war crosshairs is survival 101.

Why it matters

We think this is a necessary correction, but one loaded with landmines.

For everyday users and researchers, a less sanitized ChatGPT is an immediate upgrade. Nobody wants to pay twenty bucks a month to be scolded by a machine. When you are analyzing policy debates, writing historical fiction, or researching polarizing subjects, getting a breakdown of multiple perspectives is infinitely more useful than an unhelpful automated refusal. You do not need to change any settings; the tone shifts should filter into daily prompts naturally.

However, the strategy has an obvious flaw: the trap of false equivalence. Stripping away editorial judgment sounds great until the model treats evidence-backed facts and harmful conspiracy theories as equally valid opinions. In trying to please everyone and offend no one, ChatGPT risks turning neutrality into a refusal to acknowledge reality.

Ultimately, OpenAI's move signals the end of the hyper-cautious AI safety era. The industry is shifting from corporate scolding to user autonomy, largely driven by political self-preservation. Yet claiming to have no stance is, ironically, the most calculated political stance OpenAI could take right now.