The judge in the Musk v OpenAI^[1] case out of the northern district of California has issued an order on Musk's motion for preliminary injunction, which asked for an order temporarily stopping OpenAI from taking further steps towards its possible conversion from nonprofit to for-profit control. As might be expected for such high profile litigants, this has received quite a bit of attention in the media, pointing out that Musk's request for a preliminary injunction was denied. However, several commentators who I perceive to be interested in AI safety have jumped in to say "not so fast!". They argue that while its true that Musk was denied a preliminary injunction, digging deeper reveals the situation to be much less favorable to OpenAI then it may seem. Here's how Garrison Lovely (in the second link above) puts it:

In a narrow sense, Musk did "lose." However, as Judge Rogers notes, the bar for a preliminary injunction is extremely high. To surpass it, both the facts of the case and the legal questions involved have to clearly point in the same direction.

Therefore, it would have been extremely surprising if the judge granted the injunction. She did, however, stop just short of it.

Elaborating on the reasoning, Garrison identifies the issue as one of "standing". Rob Wiblin, the other commentator linked above, describes in a podcast interview with nonprofit legal expert Rose Chan Loui this standing issue as follows:

So in a sense, that’s a slightly boring technical legal question. Maybe the juicier interesting thing, and the even bigger update, was having demonstrated that he can bring the case of, is OpenAI doing something objectionable? Is it violating the public’s trust? And is it violating the agreement basically with Musk to use his charitable giving to pursue the charitable mission? And there the judge has seemed very sympathetic to Musk, basically.

Lovely again on the deeper implications that we can infer from this order:

So why does this ruling matter? Well, while Judge Rogers found Musk's standing uncertain at this preliminary stage, she went out of her way to signal that the core claim — that OpenAI's conversion violates its charitable purpose — could have merit if properly brought before the court.

I think part of the excitement here on the part of AI safety folks is that if they are right, this could be an important positive development and opportunity. Many people interested in AI safety (myself included) are extremely skeptical of OpenAI's actions and general approach. For people of that view, not only would this be some good news, but also a potential opportunity to make a difference. I couldn't have put it better than Lovely does, under the heading "you know who does have standing?":

Unlike Musk, the Attorneys General (AGs) in California and Delaware unquestionably have standing to challenge OpenAI's conversion

Commenters on the EA forum have started discussing writing letters the CA and DE AGs, and the possibility of filing amicus briefs in the case has also been raised there. I think the sentiment here is that the order suggests that this litigation and possible related litigation by the state AGs would be a golden opportunity for safety-concerned individuals to play a role in what OpenAI is up to.

But, are these interpretations of the order actually correct? Here, I offer an alternative interpretation of the language in the order that conflicts somewhat with that of Lovely and Wiblin.

Preliminaries

Before I dive in to the text of the order and other documents in the case to develop my own interpretation, I want to offer a couple of preliminary remarks:

Reasons

I'm going to try to go into a decent bit of detail on some of the documents in this case. Why do I feel the need to do all this? Do I desperately want OpenAI to win this case? My interest in the case in general stems from my own interest in AI safety and related issues. As with many "safety concerned" people, I have a fair deal of skepticism around OpenAI as a baseline, although perhaps not as much as some others. That said, it's not really clear to me how big of a deal either way it would be if OpenAI was ordered to halt its efforts towards conversion. It seems to me like it could vary anywhere from mildly inconvenient to catastrophic for OpenAI, and I'm not really sure how to tell where the result will actually fall on that spectrum or what the impact on AI safety would be. As a result, I don't currently have a strong opinion on whether OpenAI should or shouldn't stay under nonprofit control.

That said, I've been following the reaction of the AI safety community on this and other legal issue related to artificial intelligence, and I perceive a certain amount of overconfident takes in this general category, with an associated (I perceive) overly rosy view of litigation as a tool for change. In addition to simply wanting to ensure that the correct information (or what I believe to be the correct information anyway) is out there on this issue, I think this might also be helpful for purposes of calibrating the community's understanding of litigation as a mechanism for influencing AI policy.

Caveats

I'm not a lawyer, nor do I have any special knowledge or expertise in the law or federal courts. The pieces by Wilbin and Lovely that I reference both extensively incorporate the views of people who do have that kind of expertise. So, you might naturally ask, why the fuck would anyone trust some random pseudonymous blogger's reading of a court order over literal experts in the field?

Well, maybe you shouldn't trust my view? This obvious fact made me seriously doubt many of the things I'm about to say myself, so I couldn't blame anyone else from arriving at a similar conclusion. Perhaps these experts or some others will school me on why I'm wrong here, but I'm going to try very hard to show my work, in hopes of earning at least partial credit. I'm going to try to stick to the words of the documents relevant to the order, and to be very explicit about why I think my reading of those words is correct, or at least plausible.

Scope and links

The docket for the case is available here. As far as I can tell, all the discussion relevant here is surrounding a single part of the preliminary injunction order, section III.c., addressing breach of charitable trust. The other parts of the order clearly deny Musk's requests on standing grounds, but from my perspective the commentary addressed in this piece is focused specifically on part C. Tracing this back through the documents, this part C matches with section I.c. in Musk's motion for preliminary injunction. OpenAI's opposition to the motion for preliminary injunction is here, with section III of the argument portion addressing the charitable trust issue. This part C of Musk's motion for preliminary injunction ultimately stems from a single count in Musk's first amended complaint, count XX (20), for breach of charitable trust.

There are also two amicus briefs filed in the case. Encode Justice filed a motion for leave to file an amicus on the motion for preliminary injunction, with the proposed brief available on the same document as exhibit 1. This brief was accepted in the order here.

The second amicus brief (motion, brief), was filed by the Delaware Attorney General. This motion to file was also granted. Musk filed a motion to respond to this brief (motion, response), which was granted as part of the same order as the Encode brief.

The California Attorney General is also a defendant in this case, from my understanding because various actions by Musk have attempted to bring the AG in. The CA AG has asked to be dismissed from the case. My reading of that motion is that the CA AG is basically saying they can't be forced into being part of the case. The CA AG doesn't seem to take any positions on the substance of the case.

Is standing really the core issue?

What is "standing"?

Although I've casually encountered the idea of standing in a legal context before, since I'm going to be contesting the interpretations of actually lawyers, I figured I'd check some sources for a description of what "standing" means in a legal context.

From Cornell Law School:

Perhaps the most important element of the requirement of adverse parties may be found in the “complexities and vagaries” of the standing doctrine. “The fundamental aspect of standing is that it focuses on the party seeking to get his complaint before a federal court and not on the issues he wishes to have adjudicated.” 1 The “gist of the question of standing” is whether the party seeking relief has “alleged such a personal stake in the outcome of the controversy as to assure that concrete adverseness which sharpens the presentation of issues upon which the court so largely depends for illumination of difficult constitutional questions.” 2 This practical conception of standing has now given way to a primary emphasis upon separation of powers as the guide. “[T]he ‘case or controversy’ requirement defines with respect to the Judicial Branch the idea of separation of powers on which the Federal Government is founded. The several doctrines that have grown up to elaborate that requirement are ‘founded in concern about the proper—and properly limited—role of the courts in a democratic society.’” 3

And from the US Department of Justice:

The "case or controversy" clause of Article III of the Constitution imposes a minimal constitutional standing requirement on all litigants attempting to bring suit in federal court. In order to invoke the court's jurisdiction, the plaintiff must demonstrate, at an "irreducible minimum," that: (1) he/she has suffered a distinct and palpable injury as a result of the putatively illegal conduct of the defendant; (2) the injury is fairly traceable to the challenged conduct; and (3) it is likely to be redressed if the requested relief is granted. See Valley Forge Christian College v. Americans United for Separation of Church and State, Inc., 454 U.S. 464, 472 (1982); Gladstone, Realtors v. Village of Bellwood, 441 U.S. 91, 99 (1979); Simon v. Eastern Kentucky Welfare Rights Organization, 426 U.S. 26, 37 (1976). In addition to the constitutional requirements of Article III, courts have developed a set of prudential considerations to limit standing in federal court to prevent a plaintiff "from adjudicating 'abstract questions of wide public significance' which amount to 'generalized grievances' pervasively shared and most appropriately addressed in the representative branches." See Valley Forge, 454 U.S. at 474-75, quoting Warth v. Seldin, 422 U.S. 490, 499-500 (1975). Speculative claims that a proposed governmental action may result in injury to a plaintiff are insufficient to confer standing. See O'Shea v. Littleton, 414 U.S. 488, 497 (1974). The required injury must be both real and immediate, not conjectural or hypothetical. See Golden v. Zwickler, 394 U.S. 103, 109-10 (1969).

From this, I think there are three points about standing worth emphasizing for our purposes:

1. Standing is about whether the party bringing the suit is the appropriate party to do so. This is part of Wiblin and Lovely's point. If there is a standing issue, then what is an issue for Musk may be resolved by a different party (such as the CA and/or DE state AGs), bringing a suit instead.
2. Standing is about allegations ("alleged such a personal stake", "putatively illegal conduct"). This is a bit of a risky claim for me given that it isn't squarely stated in the quotes, I'm kind of reading between the lines, but this makes sense to me. Standing is a question of initial access to the courts at all, prior to the full adjudication of the details and facts of a case. Thus, it makes sense that it is driven by allegations of unlawful conduct, not definitive proof. Otherwise, any factual question in a case would necessarily be a standing issue, defeating the point.
3. The purpose of standing from a constitutional point of view is a separation of powers one. A court would intrude upon the domain of the other branches if it takes up issues where standing is lacking. Another way that this is sometimes put is to say that courts don't issue "advisory opinions". The purpose of Article III courts isn't to opine on matters of public policy, that is left to the other branches. Restricting the domain of courts to "cases and controversies", preserves this allocation of those questions to their proper place.

Is the relevant issue primarily one of standing?

Clearly, the issues addressed in sections A, B and D are primarily about standing, so Musk did lose several of his arguments on standing grounds. But we are concerned with the issues addressed in part C of the order. Searching the text of section C, I did not find an occurrence of the word "standing". One footnote (footnote 11) that is referenced in section C does contain the word. Here's the context in which that footnote appears:

Because the threshold question of whether a charitable trust was created remains a toss-up, Musk has not demonstrated likelihood of success on the merits sufficient to obtain an injunction. The request for an injunction barring any steps towards OpenAI’s conversion to a for-profit entity is DENIED.11 That said, given the other considerations outlined, the Court finds that the case would
benefit from an expedited trial on the claim.

And the text of the footnote:

Defendants also challenge plaintiffs’ standing. As Musk has not been directly affiliated with OpenAI for several years, any standing must come from an interest in OpenAI’s assets. California Corporations Code Section 5142(a). The Court is aware of the distinction between Restatement (Second) of Trusts § 391 and Restatement (Third) of Trusts § 94 and cmt. g, plus the California state authorities following the Restatement (Third). Thus, for purposes of this motion, the Court finds plaintiffs’ standing sufficient as a settlor given the modern trend in that direction. The motion to dismiss on this issue is DENIED. Further briefing on this topic is not necessary

I think this pretty definitively indicates that standing is not the issue that the judge is addressing in the main text of this section. Otherwise, why the "also" in the first sentence? Likewise, "for purposes of this motion, the Court finds plaintiffs’ standing sufficient" is pretty unambiguous. It's true, there is an argument about standing here, but that isn't the main issue that the court has identified. The "for purposes of this motion" part leaves it unclear to me whether this could come back later^[2], but it seems clear that the judge isn't primarily addressing standing in this section.

If the issue isn't one of standing, what is the issue? In my view, the issue is a question of fact about whether the exchanges between the parties in the early days of OpenAI were sufficient to establish some type of trust, agreement, or similar commitment:

Whether Musk’s emails and social media posts constitute a writing sufficient to constitute an actual contract or charitable trust between the parties is debatable. On the one hand, the email exchanges convey early communications regarding altruistic motives of OpenAI’s early days and even include reassurances about those motives from Altman and Brockman when they perceived Musk as upset. The Court therefore disagrees with defendants’ framing that there is no evidence that “Altman, Brockman, or OpenAI solicited his donations subject to these commitments.” (OpenAI Oppo. at 20.) At the same time, though highly suggestive, the emails do not by themselves necessarily demonstrate a likelihood of success. There is, for example, defendants’ counter-evidence implying that Musk himself considered the possibility of being the one to turn OpenAI into a for-profit. (Dkt. No. 32-17; FAC, Ex. 16.) On balance, the Court finds the emails are insufficient for purposes of the high burden required for a preliminary injunction10 and the question of likelihood of success on the merits to be a toss-up.

This is a factual question about whether some such commitment existed, and it therefore makes sense that the judge suggests a trial to resolve the factual issue. But isn't the judge saying that this factual issue goes to standing? That she needs to determine if there is an agreement in order to figure out whether Musk has standing? I don't think so (and I challenge anyone to reconcile this interpretation with the footnote). Remember, standing is about allegations. Musk clearly alleges a trust exists, and the judge seems to agree he has at least made enough of a showing in this regard such that there is a genuine disagreement between the parties about this^[3].

Why that matters

Am I just being an annoying pedant here? We've just spent all this time talking about standing and have barely mentioned the public interest, and isn't that really where all the action is? After all, several of the other parts of the injunction were denied for standing, and some of those seem like issues that one of the state AGs could raise. Even if this breach of trust stuff isn't a standing issue, isn't it essentially a technicality with regard to the stuff about public interest, which is what's really going to matter if the AGs get involved?

I don't think the issue of whether a charitable trust was created is a technicality in the view of the judge, and I think this has implications for how to correctly interpret the parts of the order that relate to "the public interest". The count of Musk's first amended complaint related to this request is breach of charitable trust. The existence of such a trust is core to what makes the alleged conduct wrongful. If there is no trust, then there is no breach, and thus no wrongful conduct. Several of the other counts in the complaint also seem to me like they could depend on facts related to this issue. There are counts about contracts, implied-contracts, and quasi-contracts. Similarly, no contract, no breach. And the fraud counts. If the statements in the early days of OpenAI by the OpenAI defendants constitute misrepresentations, then there could be fraud, but if they are not misrepresentations, then where's the fraud? Many of the counts relate at their core to what the OpenAI defendants said to Musk, and how we should interpret those statements. These aren't just technicalities, these go to what makes those counts wrongful. If you do something that you promised not to do, it fundamentally matters if you did in fact make that promise. If you misrepresented your plans or intentions to someone, it matters if your statements were in fact misrepresentations.

Now, I think many people in the AI safety community who are mad at OpenAI have the view that the fact that OpenAI started as a nonprofit and wants to convert to a for-profit is in and of itself wrongful, or at least very close to it. The excitement about this court order among that group I think comes from a perception that the judge is substantially agreeing with that perspective. I'm not so convinced that is the case.

Conditionality

What is the judge communicating?

First, I think its worth noting, the interpretation of Wiblin and Lovely is a bit odd in light of what I said above about standing. Remember, courts aren't in the business of issuing "advisory opinions". A district court is not supposed to be "adjudicating 'abstract questions of wide public significance' which amount to 'generalized grievances' pervasively shared and most appropriately addressed in the representative branches". Here is what Lovely suggests about what the judge is trying to do with this order:

So why does this ruling matter? Well, while Judge Rogers found Musk's standing uncertain at this preliminary stage, she went out of her way to signal that the core claim — that OpenAI's conversion violates its charitable purpose — could have merit if properly brought before the court.

I'm not sure if going out of your way to signal things when you are "uncertain" about standing is all that consistent with the ideas about standing and the role of courts quoted above.

Here is an exchange from the 80k hours podcast describing the situation:

Rose Chan Loui: Yes. So I think that is the encouraging part. If we can get past this hurdle, the judge is indeed very sympathetic and concerned about what happens to these assets. And like we were talking about earlier, is this a signal to other parties, mainly to the attorneys general?

Rob Wiblin: Does she clarify what she thinks is probably objectionable about what OpenAI is trying to do?

Rose Chan Loui: Well, not really. I mean, there’s a part that to me doesn’t really make sense where they talk about the last argument with the federal tax benefits and kind of make a big deal of it.

Chan Loui then goes on to describe some issues raised in a part of the order that I'll bring up later which I also found quite confusing, but I want to focus for the moment on that question, what does the judge find objectionable? Several commentators so far seem very convinced that the judges is trying to send signals about bad conduct she thinks has occurred, but its interesting that there doesn't seem to be a ton of clarity about what that bad conduct actually is.

Why comment on the public interest?

If the judge isn't trying to send some type of message, why all this writing about the public interest in the order at all? Simply, its one of the factors established by precedent that a judge is supposed to consider when deciding whether to grant a preliminary injunction:

The United States Supreme Court has consistently held that a “preliminary injunction is an extraordinary remedy never awarded as of right.” Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7, 24 (2008). A plaintiff seeking a preliminary injunction must establish that the plaintiff is both likely to succeed on the merits, and likely to suffer irreparable harm in the absence of preliminary relief, and further that the balance of equities tips in plaintiff’s favor, and that an injunction is in the public interest. Id. at 20. Alternatively, a preliminary injunction may issue where “serious questions going to the merits were raised and the balance of hardships tips sharply in plaintiff’s favor” if the plaintiff “also shows that there is a likelihood of irreparable injury and that the injunction is in the public interest.” All. for the Wild Rockies v. Cottrell, 632 F.3d 1127, 1135 (9th Cir. 2011). This is the Ninth Circuit’s “sliding scale” approach, in which “the elements of a preliminary injunction test are balanced, so that a stronger showing of one element may offset a weaker showing of another.”

4 factors to consider, likelihood of success on the merits, public interest, balance of the equities, and irreparable harm. In the Ninth Circuit, these factors can trade off, so weakness in one can be compensated for by strength in another. On the likelihood of success on the merits question the judge found this to be a "toss-up". Thus, in order to resolve the preliminary injunction question, the judge must look to the other factors, which all kind of bleed together into what would colloquially be called "public interest".

So, with that out of the way, what is my view on what the discussion of the public interest means, and what it is that the judge thinks is objectionable here?

What the order says

I obviously can't say what Judge Rogers' internal deliberations are like. It's definitely possible that her ruling is colored somewhat by her impressions of the parties, their conduct, and her general views about the goodness or badness of what OpenAI is doing or plans to do. But it's my view that textually and legally, the thing that Judge Rogers is saying is objectionable in the order is in fact exactly what I said above, the breach of charitable trust itself is the bad conduct being addressed. From the order (emphasis in original):

On balance, the Court finds the emails are insufficient for purposes of the high burden required for a preliminary injunction10 and the question of likelihood of success on the merits to be a toss-up.

Because of this, the remaining Winter factors are derivative of, and dependent on, the first. If a trust was indeed created, then preventing or remedying breach would be in the public interest.

"If a trust was indeed created". The other factors (public interest, balance of equities, irreparable harm) are "derivative of, and dependent on" the determination of whether "a trust was indeed created". What would be in the public interest? "Preventing or remedying breach". That's public interest, what about the balance of the equities? The order (emphasis in original):

Next, and similarly, if a trust was created, the balance of equities would certainly tip towards plaintiffs in the context of a breach.

"If a trust was indeed created", "in the context of a breach", conditional on those circumstances "the balance of equities would certainly tip towards plaintiffs". The order continues (emphasis in original):

As Altman and Brockman made foundational commitments foreswearing any intent to use OpenAI as a vehicle to enrich themselves, the Court finds no inequity in an injunction that seeks to preserve the status quo of OpenAI’s corporate form as long as the process proceeds in an expedited manner.

There would be "no inequity" if the conversion where halted "to preserve the status quo", "as long as the process proceeds in an expedited manner". I think this is saying something much more narrow than Wiblin and Lovely seem to think. If OpenAI did in fact agree to avoid converting to a for-profit always and forever, how could it possibly be contrary to the balance of equities to make them simply do what they already agreed to do from now until trial? Assuming there was a trust created they already agreed not to do it! But the judge's point here isn't to signal about the public interest, it's that the direction these other preliminary injunction factors weigh is expressly conditional on the first factor, which boils down to the question "was a charitable trust created".

Coming to the third factor, this is where I think maybe people of the view that the judge is expressing sympathy with Musk have the most to stand on, but it's also the most confusing to me:

The third element is likewise similar. Plaintiffs attest that the irreparable harm in this instance flows from the abuse of donor funds and tax deductions. Plaintiffs first described the harm as being the “competitive advantage . . . of being able to raise . . . publicly tax-subsidized money to develop technology, and then transferring the technology to an opaque web of for-profit entities and benefiting from that when competitors are supposed to compete on a level playing field.” (Dkt. No. 109 at 64:6-13.) When the Court paraphrased and asked if it was accurate to say the claim of irreparable harm is “that people shouldn't be able to take a tax deduction from a nonprofit who then can convert that money for for-profit enterprises,” plaintiffs agreed. (Id. at 64:16-21.) The Court agrees that significant and irreparable harm is incurred when the public’s money is used to fund a non-profit’s conversion into a for-profit.

The first thing that confuses me is that my understanding of "irreversible harm" is that is supposed to be "not money". If the defendant can just pay you, that's reversible. This all sounds like "money", to me. Maybe the original part about "competitive advantage" could be "not money", but then, to me, it seems like the judge's summary kind of changes the meaning there to be essentially "money".

That said, on a practical level, it does seem like it would be really hard to unwind a conversion that had already happened. And if this factor were the only one in question, I think the judge would probably still deny the motion anyway, so if it's not a dispositive issue, it kind of makes sense to keep options open. The judge could always decide later that money is a sufficient remedy.

But on the issue of what this means in terms of the judge's view of the conduct here "The Court agrees that significant and irreparable harm is incurred when the public’s money is used to fund a non-profit’s conversion into a for-profit", to me seems like the quote that is closest to the interpretation of Wiblin and Lovely, and certainly does seem to suggest the judge might have serious questions about the conversion. But at the same time this entire discussion is in the context of the judge saying all three factors are dependent on the charitable trust questions. Even at the beginning of this section, "the third element is likewise similar". So my interpretation is that the judge still views that as the central question, just that if that resolves in favor of Musk, there is a real question of harm here.

Conclusion

So, in conclusion, I think what the judge is saying is quite a bit narrower than what others are claiming. Despite that, I do think there is a big part here that is a loss for OpenAI, which Lovely also points out:

"This is a big win for Musk," says Michael Dorff, the executive director of the Lowell Milken Institute for Business, Law, and Policy at UCLA. "Even though he didn't get the preliminary injunction, the fact that there is a pending trial on this issue and that his claim wasn't denied is a pretty big impediment to [OpenAI] moving forward expeditiously," he says.

OpenAI probably was hoping to have the entire case dismissed. Even if a lot of the counts get dismissed for standing, it looks like they will still have to go to trial. I think for a lot of defendants, and probably OpenAI here, the case getting past a motion to dismiss and going to trial on anything is basically a loss. In that sense, this is indeed bad news for OpenAI.

Position of DE Attorney General

I don't have any arguments or recommendations for the following, but I thought it might be worth highlighting some information that could be relevant to how this and related matters are proceeding. The amicus brief from the DE AG goes over some information relevant to how that office is thinking about this stuff. They introduce their reason for filing the brief (the "Proposed Transaction" refers to the possible conversion of OpenAI):

The Delaware Attorney General submits this brief to make the Court aware that she is conducting a concurrent review of the Proposed Transaction on behalf of the public pursuant to her authority under Delaware law. Specifically, the Delaware Attorney General has authority to review the Proposed Transaction for compliance with Delaware law by ensuring, among other things, that the Proposed Transaction accords with OpenAI’s charitable purpose and the fiduciary duties of OpenAI’s board of directors. The Delaware Attorney General has not yet concluded her review or reached any conclusions regarding the Proposed Transaction, and she takes no position on the propriety of a preliminary injunction in this case. The Court, however, should be aware that Delaware will not hesitate to take appropriate action to protect the public interest if the Delaware Attorney General’s review demonstrates that such action is warranted.

There are several places where the DE AG seems to highlight their important role:

The Delaware Attorney General has the authority and responsibility to oversee Delaware charitable corporations.

And more:

Pursuant to her responsibility and authority to protect the public by supervising and overseeing charities, the Delaware Attorney General is in ongoing dialogue with OpenAI regarding the Proposed Transaction. The Delaware Attorney General’s authority includes the ability to review the transaction for, inter alia, whether the charitable purpose of Open AI’s charitable assets would be lost or impaired; whether any intended for-profit entity will adhere to the existing charitable purpose of OpenAI (assuming the not-for-profit entity remains affiliated with it); whether OpenAI’s directors are meeting their fiduciary duties; and whether, if applicable, the Proposed Transaction satisfies Delaware’s “entire fairness” test.

The DE AG also elaborates somewhat on what actions they are currently taking:

The Delaware Attorney General is reviewing the Proposed Transaction to ensure that OpenAI is adhering to its specific charitable purposes for the benefit of the public beneficiaries, as opposed to the commercial or private interests of OpenAI’s directors or partners. “Any action that poses a palpable and identifiable threat to those [charitable] goals, or that jeopardizes its assets would be contrary to the Certificate [of Incorporation] and hence ultra vires.” Oberly, 592 A.2d at 462.

The AG also makes the case that they are in a good position to make these types of determinations:

The Proposed Transaction—or its enjoinder—is likely to have
multiple and profound impacts on the public. The Delaware Attorney General is well-positioned to consider how the interests of the charity’s beneficiaries are best served. If the Delaware Attorney General concludes that the Proposed Transaction is not consistent with OpenAI’s mission and its obligations to the public, that OpenAI’s board members are not fulfilling their fiduciary duties, or, if applicable, that the value of the Proposed Transaction or the process for deriving it is not entirely fair, the Delaware Attorney General will take appropriate action, whether through a requested modification of the terms of the Proposed Transaction or, if necessary, by seeking an injunction if and when she determines judicial intervention is necessary.

The DE AG takes no position in the brief as to the preliminary injunction, instead the purpose is simply to inform The Court of what the DE AG is doing. I found it interesting the way that the role of the AG is highlighting. Reading the brief made me wonder if perhaps the DE AG is trying to send a signal of their own, about who the ideal decision makers on these issues might be.

Musk seems to have similar thoughts. From Musk's response:

The Delaware Attorney General’s amicus brief is correct that OpenAI’s unprecedented conduct in operating a charity for private gain is a matter of “profound” importance. Del. Amicus at 4.1 However, its implicit suggestion that the office’s review of the “Proposed Transaction” is enough to protect the public interest misapprehends the issues at hand. Delaware does not, because it cannot, purport to protect the interests of Elon Musk, Shivon Zilis, and competitors like xAI, none of whom are Delaware citizens. Delaware does not, because it cannot, purport to regulate Microsoft, Reid Hoffman, or Deannah Templeton, none of whom are Delaware citizens. And Delaware does not, because it cannot, purport to have special competence in protecting the federal interests embodied in federal law.

I think its worth keeping in mind that (at least in my opinion), there is a real possibility that the state AGs will try to resolve their concerns directly with OpenAI, without necessarily going to litigation.

1. ^{^}

   Although there are several other parties in this case on both sides, I plan to refer to the case as "Musk v OpenAI", the plaintiffs as "Musk", and the defendants as "OpenAI".

2. ^{^}

   But in that case, why "The motion to dismiss on this issue is DENIED"?

3. ^{^}

   That leaves open the questions raised by OpenAI about standing referenced above (reversionary interest, Restatement (Second) of Trusts vs Restatement (Third) of Trusts). I'm not entirely sure what the status of that is or what the issues are, but it doesn't seem to me like that is the central thrust of what is being discussed by the judge here, so I'm going to leave it to the side for the moment.