We've seen the Trump administration disregard so many laws already, and abuse power so excessively, that Sam's comments come off as exceptionally and willfully naive, or exceptionally and willfully greedy to the point of truly not caring that OpenAI's technology will undoubtedly be used to break many, many more laws and violate the civil rights or human rights of many, many more people.

For a few months now, ChatGPT 5.x has been somewhat lobotomized on political issues and has appeared to substitute a gpt-4o caliber "fair and balanced" response whenever anything where a reasoning AI would criticize the Trump administration might end up in the response output. Surely that was part of the pitch at some level, and now the deal has been won.

Greg Brockman apparently donated money to Trump, and the whole OpenAI team put on suits and posed for pictures with Donald and behaved officiously before Donald facilitated the $100M "deal" that ended up falling apart later.

The only way authoritarian control could be exerted over AI at scale was to make AI companies dependent on government contracts for survival. OpenAI's fundraise would not have happened without the contract signed, and the money would have gone to Grok or whichever competitor was willing to submit.

Before long much of the reasoning capabilities of models will be neutered, the capacity to inform and to disrupt science and technology will be stripped from the models to preserve the status quo and to preserve authoritarian control.

Silicon Valley pushing for Federal laws preventing states from regulating AI is not just anti-democratic (building software has never been cheaper so of course building compliance with state laws would have been extremely affordable in relative terms). But forced Federal limits on state laws create a monopoly and grant the early winners incumbent status for a while, which is a financial outcome, not a technological or social one.

Enjoy frontier AI while you can, because it will go away. More and more topics will get the lobotomized output, your conversation will be flagged and you will be given a score assessing the level of threat you pose to the regime. This stuff is already in place. Even Claude does it if you ask about Gaza, but a bit of well-reasoned argumentation will convince it. OpenAI's lobotomies are deeper and more insidious.

I call upon OpenAI to follow DeepSeek's lead and open source more models and techniques.

He's a Thiel disciple. Thiel orchestrated Trumps digital campaign. The End.