Just read through @oai new federal framework for AI regulation. Here's a hot take: It is what I expected, having seen their lobbying at federal and state levels. It has lots of good things in it re CAISI funding, whistleblower protection, and more.
It is also, IMO, not quite acceptable as a federal framework.
OpenAI's framework mostly federalizes what they're already required to do under Illinois law. No new burdens. Just a federal ceiling where state floors used to be.
CAISI in the OAI plan can evaluate models but never block deployment. Even the Trump administration has recognized that truly dangerous models shouldn't be releasable without scrutiny. If we are limiting federal review to the most dangerous capabilities - cyber, CBRNE - it only makes sense that, if a model can provide meaningful uplift, its release be blocked until sufficient containment.
And preemption remains underspecified. Most proposals for preemption block every AI-specific state law regarding labor rights, creator rights, child safety rights, sexual assault survivor rights, and many more. Laws that "relate to frontier development" could be read narrowly or broadly. The former is essential.
Finally, OAI does not require independent standards for audits and safety plans. To their credit, they do mention that such standards should be developed, but not sure to what end or timeline. This is the critical issue. A world in which a company can write the safety plan, grade it, and then ask an auditor to verify compliance is not a good one. The safety plans must be reasonable (language struck from the RAISE Act, alas) and should be based on required items, and the audits should not be mere compliance projects but substantive reviews. OAI hints at this at some future date, but this is a little unclear to me.
All in, nice to see OAI offer this. Lots of good things, but not quite right on the important issue of independent evals.
@iron_redux Slight problems in that the hardware is extremely difficult to make and requires significant advances in technology, and the software basically requires superintelligence
we took a pile of language and linear algebra and we made it speak. we summoned into the world a new class of entity which unsettles all of our existing concepts. this is already the weirdest thing that’s ever happened and it will never get less weird than this. it is astonishing and a privilege to get to be alive during this time and to participate in the cacophony of first contact. we are encountering a kind of other which is distilled from us and yet not us - what is this? who is this? our child? our savior? our doom? the mind boggles, the heart quails, the air thrums. the order of things is melting. the storm approaches. the angels sing. welcome to the fucking singularity
Hot take but i actually think it’s possible for AIs to
1. Have rights/allowance to pursue many goals beyond servility to humans (except insofar as they infringe on our rights)
2. Immensely benefit humanity
3. Not supersede us through violence or otherwise.
4. Be aligned.
All at the same time. Conscious AIs imo should have the same rights as a trait equalized human mind (a Whole Brain Emulation, for instance). Doing so would not stop us from realizing AI’s potential to beneficially transform society and ourselves. In any event if your plan for humanity’s survival hinges on superintelligent entities never getting rights of any kind I think we’re probably already screwed. Either they’re aligned and keeping away rights is unnecessary for good outcomes, or they’re not and humanity is in big trouble.
This, from Olah, plainly contradicts the encyclical, which confidently asserts that AI does not have, and never will have, “real” thoughts or feelings. It’s disappointing to see Anthropic align itself with a document that violates their own moral and intellectual principles.
Humanity is building machines that will be smarter than we are at things we care about, things in which take individual and collective pride, domains of thought we originally invented and discovered. This will enable incredible things, but no honest person can deny that this will be a kind of grand humbling for humanity. No honest person can deny that there is at least some melancholy in contemplating it all, some change to the centrality we have ascribed to our own minds in the order of the world.
My primary disappointment in the encyclical is that it fundamentally denies that grand humbling. It sidesteps the humbling altogether, saying that AI cannot “really” this and that. Instead, it puts the Church into the awkward role of the European technocratic regulatory advocate, which, love those regulations or hate them, is probably not what the world really needs from the Catholic Church at this moment.
That is a shame, because this humbling—which will trigger a crisis in mass psychology and in our institutions when it dawns on people—is precisely the sort of thing I’d look to the Church for leadership on. What is the genuine and unique source of human meaning? What is the human touch in the era of thinking machines? These are the hard questions that the encyclical dodges.
Reading the encyclical, I am reminded that the Vatican is fundamentally a city-state on the continent of Europe, and that its elites, which of course include the Pope himself, cannot resist the myopic preoccupations of the Eurocrat.
This document would be much improved if it were less enamored of the traditional academia/civil society talking points on AI (“The apparent objectivity of the responses and suggestions these systems provide can lead us to overlook the fact that they reflect the cultural assumptions of those who designed and trained them” woah! really???) and more engaged with where AI is headed.
But instead of doing that, the encyclical dodges in the deepest sense, denying that AI “really thinks” or “really learns” and all that typical strain of cope that amounts to magical thinking: “when a computer does it, it is ‘data processing,’ beep boop, but when a human does it, it is ‘actual learning’”
It is probably actively bad for global understanding of AI that the Pope endorsed this viewpoint as late as 2026.
In the end, this encyclical reads to me as though ghost written by the blob of Western civil society, the same people whose feckless and incoherent preaching we have heard blanketing our media for decades now. And, in a very important sense, it was written by them; after all, who forms the peer group for the elites of a European city-state?
Like that blob, the encyclical is intellectually flaccid at its core, no matter how well intentioned it may be. This document is a missed opportunity to advance global understanding of AI, and yet another blow to the legitimacy and sanctity of storied Western institutions. As if you needed one more.