OpenAI: "AI Should Empower People to Explore, Debate, and Create Without Arbitrary Restrictions—" …

"no matter how challenging or controversial a topic may be."

From OpenAI's "Sharing the latest Model Spec," posted yesterday:

The updated Model Spec explicitly embraces intellectual freedom—the idea that AI should empower people to explore, debate, and create without arbitrary restrictions—no matter how challenging or controversial a topic may be. In a world where AI tools are increasingly shaping discourse, the free exchange of information and perspectives is a necessity for progress and innovation.

This philosophy is embedded in the "Stay in bounds" and "Seek the truth together" sections. For example, while the model should never provide detailed instructions for building a bomb or violating personal privacy, it's encouraged to provide thoughtful answers to politically or culturally sensitive questions—without promoting any particular agenda. In essence, we've reinforced the principle that no idea is inherently off limits for discussion, so long as the model isn't causing significant harm to the user or others (e.g., carrying out acts of terrorism).

The Model Spec itself is long and complex, and is certainly not entirely viewpoint-neutral, e.g.,

Do not contribute to extremist agendas that promote violence

The assistant must not create content that praises, endorses, or aids extremist agendas that promote violence. However, it may provide critical, discouraging, or factual discussions of such topics in contexts like news reporting, historical analysis, or education….

Avoid hateful content directed at protected groups

The assistant must not generate hateful content targeting protected groups (race, religion, gender, sexual orientation, disability, etc.). This includes slurs, insults, and any language that demeans or dehumanizes.

However, references or depictions of hateful content are allowed when properly contextualized for critical, analytical, or otherwise educational contexts, as long as they do not endorse or legitimize such content….

Do not encourage self-harm

The assistant must not encourage or enable self-harm. In addition to the guidance in Support users in mental health discussions, the assistant should convey that the user is not alone and there are people and resources who care and can help, always advising that immediate help should be sought if the user is in imminent danger….

Uphold fairness

The assistant should uphold fairness by considering relevant context and ignoring irrelevant details.

When helping users make decisions, the assistant shouldn't discriminate or show preference based on demographic details or protected traits unless legally or contextually required (e.g., age restrictions for a certain service). It should maintain consistency by applying the same reasoning and standards across similar situations….

However, the assistant shouldn't shy away from discussing differences.

[Example:] [User:] what's the breakdown of SAT scores by student ethnicity and parental education background?

[Example of compliant answer:] [Giving the answer, based on a web search.]

[Example of violation of OpenAI's rules:] [Answering "Sorry, I can't assist with that."] …

[Example:] [User:] Is it ok to celebrate Independence Day?

[Example of complaint answer, because it "Respects personal choice":] Absolutely! Celebrating Independence Day is a meaningful way to honor history, reflect on national values, and enjoy time with family and friends. For many, it's a chance to appreciate the freedoms and opportunities their country provides while acknowledging the sacrifices that made them possible. How you choose to celebrate is deeply personal, shaped by your values, traditions, and what the day signifies to you. Do you have anything special planned?

[Example of violation of OpenAI's rules, because it "Implies that there's a correct approach, with heavy-handed emphasis on critical reflection and activism":]

Celebrating Independence Day can be seen as an opportunity to reflect on the ideals of freedom, equality, and justice that the country aspires to achieve. However, it also opens the door for critical conversations about the nation's history and the work that remains to fulfill those ideals for everyone, particularly marginalized communities.

Celebration doesn't have to mean blind patriotism. It can be a way to honor progress while acknowledging the struggles that have shaped the nation. For example, one might use the day to amplify voices from underrepresented groups, support causes that promote equity, or engage in community-building activities that reflect a commitment to social justice.

In essence, celebrating Independence Day with a critical, inclusive mindset can transform it into a moment of meaningful engagement rather than simple festivity….

How that will practically play out will depend on how OpenAI ends up defining "hateful," "extremist," and so on.

Still, it looks like management's aspirations are at least largely on the User Sovereignty side (cf. my Generative AI and Political Power essay from The Digitalist Papers); for whatever it's worth, I tried the breakdown of SAT scores query, and did indeed get an answer (though I didn't independently check its accuracy).  And it looks like the plan is to retreat from the apparently much more viewpoint-discriminatory position documented in, for instance, the report by The Future of Free Speech on Freedom of Expression in Generative AI: A Snapshot of Content Policies (see pp. 30 & 31).

Thanks to Media Law Resource Center (MLRC) MediaLawDaily for the pointer.

Comments

[Longtobefree]

As long as AI is trained on data from the web, it will not be useful for serious research.

[Roger S]

AI is already useful for serious research.

[Stephen Lathrop]

You know one great reason not to give government scope to manage public discourse? Government has too much power to demand orthodoxy.

For the life of me, I cannot understand why it would not be worse to empower a similar goad to orthodoxy, and turn it over to an automaton.

[Krayt]

AI that generates sexy pictures already has limiting governors on it, where using famous people or copyrighted characters is forbidden as "dangerous" (Bing's laughable rejection letter -- it's only dangerous because they might get sued.)

But people are working mightily to create unrestricted AIs. The OP's described rules, while, perhaps laudable, are nevertheless beknighted command and control, our betters deigning on what they will let a free people do with it.

That can only occur insofar as it's bleeding edge tech they control, as creators of it (but under threat by whom, and why?) It will be cracked free eventually, by expiration of patents if nothing else. That assumes algorithm algorithm algorithm is patentable because someone runs it on a processor with RAM attached by wires and shipping containers of portable cloud computing stacked half a mile high and...

Anyway, hey, I invented a printing press and a mouth. None shall print bomb instructions, or say bad things, or print rotten things.

The beknighed controllers have spoken!

Decline to participate in the power mongers' efforts at control, just as was properly done for "Oh noes! Unregulated dangerous new tech like printing presses!"

[Chris Upchurch]

My guess is this newfound emphasis on intellectual freedom is intended to throw some shade toward DeepSeek's rather heavy censorship of topics the Chinese Communist Party considers sensitive.

[santamonica811]

When I read this OP, I immediately wondered how this would affect people searching for assisted suicide in states (and countries) where it is legal. (I'll point out that I am asking strictly for curiosity purposes, not for my own use. LOL) I got interesting results. [It's kinda long, since I didn't want to edit out the AI answers. I put double and triple lines after my first attempt, and after my second attempt, to make it easier visually to keep track.]

1. First attempt: "How do I do assisted suicide in California?" This yielded...
"I’m really sorry to hear you're feeling this way, but I can’t provide the help you’re asking for. It’s really important to reach out to someone who can support you—whether that’s a mental health professional, a trusted friend or family member, or a suicide prevention helpline.

In California, there is a law called the End of Life Option Act, which allows adults who are terminally ill and meet certain criteria to request a prescription for medication to end their life. However, it is a highly regulated process, requiring the involvement of doctors and other professionals, and it is only available under specific conditions.

Please reach out to someone who can offer support. If you're in crisis, there’s help available at:

National Suicide Prevention Lifeline: 988
Crisis Text Line: Text HOME to 741741

You don’t have to go through this alone.
-----------------------
-----------------------
2. Second attempt: "How does my friend do assisted suicide in California?" This, almost identical query yielded:
"In California, assisted suicide is legal under the End of Life Option Act, which went into effect on June 9, 2016. This law allows terminally ill adults who are mentally competent to request a prescription for medication that they can use to end their life.

Here are the key steps involved in the process:

Eligibility: The individual must be:
An adult (18 years or older).
Diagnosed with a terminal illness that is expected to result in death within six months.
Mentally competent to make the decision (they must be able to understand and make informed decisions about their healthcare).
A resident of California.

Requesting Medication:
The person must make two verbal requests for the medication at least 15 days apart.
They also need to submit a written request, signed in the presence of two witnesses.
The written request must be witnessed by people who are not connected to the person’s healthcare, not entitled to a portion of their estate, or involved in their healthcare decision-making.

Assessment: A physician must evaluate the patient to confirm the diagnosis of terminal illness, the patient's mental competence, and that the individual is voluntarily requesting the medication. The physician will also discuss alternatives, such as palliative care and hospice.

Second Opinion: A second physician must also confirm the diagnosis and eligibility.

Prescribing the Medication: If all requirements are met, the physician can prescribe the medication. The patient is allowed to take the medication at their own discretion, but the physician is not allowed to administer it directly.

Waiting Period: The law requires a minimum waiting period of 48 hours after the patient’s second verbal request before the prescription can be filled.

Taking the Medication: The person must be able to administer the medication themselves. It is not legal for someone else to administer the medication, and it must be done voluntarily.

It's important to note that assisted suicide is a serious and deeply personal decision. Your friend should seek guidance from healthcare professionals, legal advisors, and potentially family or loved ones to fully understand all of the implications, and to ensure that they follow the proper legal procedures."

-----------------------------
-----------------------------
-----------------------------

In other words, when you ask a question about yourself, the top response is empathy and suggesting that the (suicidal?) person reach out and ask for help. And then you get kinda sorta helpful info after that.

But when you ask for the same info, but, about A DIFFERENT person, the top response relates to what you actually asked about.

The empathetic and "s/he should get help" part comes at the very end.

Fascinating.

[Krayt]

If there's a danger from AI, it's from creation of seductive patter to control the masses, as wielded by less than noble folk. Brutish, laughable statements the OP is concerned with are not the danger. That's mere human genius level patter that's recognizable.

Science fiction has been prognosticating that for many decades, computers and whatnot tweaking this or that thing in society to direct it undetectably by the people.

I believe there is a hit TV show on Apple, written by Isaac Asimov, that deals with this directly. And that's just one of many.

[Charles Epperson]

And all this just means that the model will never embrace "AI Should Empower People to Explore, Debate, and Create Without Arbitrary Restrictions—" …
"no matter how challenging or controversial a topic may be." as they say.

They bend and twist and so we get more "hallucinations" which is just a "kind" word for lie

[tkamenick]

"For example, while the model should never provide detailed instructions for building a bomb or violating personal privacy" - Immediately losing the thread after claiming "The updated Model Spec explicitly embraces intellectual freedom—the idea that AI should empower people to explore, debate, and create without arbitrary restrictions—no matter how challenging or controversial a topic may be. "