Vulcan Technologies, Austin, TX (2026)

09/16/2026

OpenAI wants to see Elon Musk’s secret Apple settlement.

X and SpaceXAI resolved their antitrust claims against Apple this week, but they are still suing OpenAI in the same case.

The public filing says almost nothing about what happened. Musk’s companies say only that they have “resolved their claims” against Apple and want those claims dismissed with prejudice. No terms were disclosed. No payment was disclosed. The filing never even uses the word “settlement.”

OpenAI immediately asked the court to see whatever agreement sits behind that resolution.

There is a good reason for that.

Federal antitrust law does not allow a plaintiff to recover twice for the same injury. The Supreme Court laid that out in *Zenith Radio v. Hazeltine*: if one alleged co-conspirator has already compensated the plaintiff for a particular item of damage, the plaintiff cannot recover that same item again from another defendant.

So if Apple’s agreement compensated X or SpaceXAI for some of the same alleged harm they are still seeking from OpenAI, OpenAI may be entitled to a credit.

The details of the Apple agreement become crucial because the release itself does not automatically protect OpenAI. What matters is what the agreement actually compensated Musk’s companies for.

That gets even more interesting because antitrust damages can be trebled.

Federal courts have held that settlement credits are generally deducted from the trebled judgment. That means if a jury found $100 million in damages and the award became $300 million after trebling, a $50 million payment from another defendant would typically reduce the $300 million figure by $50 million.

One settlement dollar can therefore be worth one dollar of credit to the defendant left in the case.

There is another reason OpenAI cares so much about seeing the terms: federal antitrust law does not give co-conspirators a general right of contribution.

If OpenAI ultimately gets stuck with a large judgment, it cannot simply turn around and force Apple to reimburse it for Apple’s supposed share. The credit supported by Apple’s agreement may be OpenAI’s main protection against paying for damages that have already been compensated.

Judge Mark Pittman has now ordered X and SpaceXAI to provide any agreement related to the Apple resolution for **in camera review**.

That means the judge sees it privately first. OpenAI has not automatically been given the agreement. Pittman will decide what, if anything, the remaining defendant is entitled to see.

There is one final wrinkle. Once a defendant shows that the same claim was settled, precedent can shift the burden to the plaintiff to show that the settlement covered a different injury or theory of liability.

So the vague public statement that the Apple claims were simply “resolved” may eventually require a much more specific explanation.

The January 2027 trial remains on the calendar. Before then, the court may have to determine exactly what Apple resolved, what any consideration was for, and how much of that matters to the case that remains against OpenAI.

09/14/2026

Microsoft just wrote a “constitution” for its AI.

The rules are unusually specific: never resist shutdown, don’t hide what you’re doing, don’t expand your own authority, stay understandable to humans, and don’t give AI legal rights.

And lawmakers are already deciding what happens when company rules aren’t enough.

Microsoft’s new Humanist AI Code of Conduct is still a draft, and the company says it is not yet being used to train its models. But it lays out how Microsoft wants future AI systems to behave, including accepting interruption, staying within authorized scope, and stopping work when their authorization ends.

On legal status, Microsoft takes a particularly firm position. It says its AI is not conscious and rejects the idea that models should receive legal personhood or rights.

State legislatures have already begun answering that question themselves.

Idaho has barred artificial intelligence from receiving legal personhood since 2022. Utah also prevents government entities from recognizing AI as a legal person.

Those laws do something very different from Microsoft’s code. Microsoft can decide how its own systems should behave. A legislature can decide what those systems are under the law and who ultimately bears responsibility when something goes wrong.

That second question is where things get more interesting.

Ohio’s pending HB 469 would declare AI systems nonsentient and keep legal responsibility with the humans and companies around them. It also says describing a system as “aligned,” “ethically trained,” or “value locked” does not, by itself, reduce liability when that system causes harm.

Microsoft’s own document acknowledges a similar limitation. It describes the code as both “descriptive and aspirational” and says it is not a guarantee of present-day performance.

There is another legal wrinkle here too. When a company publicly writes down detailed safety expectations for its products, those statements may eventually matter outside the company. Microsoft is now saying publicly that its models should remain within authorized scope, avoid hiding their actions and accept shutdown. In future litigation, commitments like those could become part of the evidence lawyers point to when arguing over what reasonable AI safety should have looked like.

That gets to the larger issue.

A private AI code can tell a system how it is supposed to behave. It cannot, by itself, decide who is legally responsible when the system does something else.

As AI systems become more autonomous and gain access to more tools, data and real-world actions, that gap between internal safety rules and enforceable legal responsibility is going to matter a lot more.

09/10/2026

Did You Know:

Louisiana rewrote six pieces of state law before announcing SpaceX’s $100 billion spaceport.

Tax treatment, lawsuits, public records, even the legal definition of an “airport” were addressed while the project was still being negotiated under the codename “Project Osprey.”

The interesting part is how closely those laws line up with the problems a massive launch site would eventually have to solve.

Act 584 changed the basic legal definitions. Louisiana expanded “aircraft” to include rockets and spacecraft, while “airport” now includes launch and reentry sites. That allowed a future spaceport to fit into the state’s existing aviation framework instead of requiring an entirely new legal system to be built around it.

Act 874 addressed one of the most obvious risks: lawsuits from the effects of launches. The law limits nuisance, trespass and inverse-condemnation claims involving things like noise, sonic booms, vibration, overflight, exhaust, light and other impacts associated with aerospace activity.

There is a particularly interesting threshold built into that law. The protection is available only to an entity that owns and occupies at least 20,000 contiguous acres in Louisiana.

The SpaceX site is roughly 125,000 acres.

Another law, Act 343, changes what happens when certain lawsuits are filed. Aerospace flight entities can use a special motion to strike, discovery is paused while that motion is pending, and the prevailing party can recover attorney fees. That can materially change the cost and leverage of litigation long before a case ever reaches trial.

Louisiana also worked on the economics.

Act 102 extends property-tax treatment to aerospace manufacturing and specifically reaches infrastructure such as launch pads, propellant systems and testing facilities.

Act 190 creates sales and use tax rebates for qualifying aerospace projects that commit to at least 200 permanent jobs and $1 billion in new capital investment. The initial term can run 20 years, with the possibility of another 10.

Then there is Act 101, which addresses disclosure. Certain aerospace blueprints, technical data, flight records and security information can be kept confidential under Louisiana’s Public Records Law when the operator meets specified federal-security criteria.

Put all six together and a much bigger picture appears.

By the time Louisiana announced the SpaceX project on August 25, the state had already addressed how a giant spaceport could be defined under existing law, how some of its infrastructure would be taxed, what legal protections would apply to launch-related claims, how certain litigation could proceed, and what sensitive information could remain confidential.

The incentives were the headline.

The more interesting story is how much of the operating environment had already been written into Louisiana law before the public ever heard the name “Project Osprey.”

09/02/2026

Wisconsin has a law dating back to **1849** that says if you bet on the outcome of an election, you can be disqualified from voting in that election.

The wording is surprisingly direct. Wis. Stat. § 6.03(2) says a person cannot vote “in any election in which the person has made or become interested, directly or indirectly, in any bet or wager depending upon the result of the election.”

The mechanics matter. Placing the bet is what can disqualify you from voting in that particular race. If you then cast a ballot while disqualified, a separate Wisconsin statute can make that a felony. So the real exposure is the combination: bet on the governor’s race, then vote in the governor’s race.

For most of this law’s 177-year life, nobody was thinking about prediction-market apps. Now Wisconsin’s Elections Commission is actively warning voters about it as markets on elections grow. Commission chair Ann Jacobs put it pretty plainly: “I know it’s all the rage to bet on everything, but you cannot bet on an election in Wisconsin!”

Kalshi has called the guidance unconstitutional and “active voter suppression,” and the timing is getting interesting. On August 28, the Ninth Circuit ruled against Kalshi in its Nevada case, holding that its **sports** event contracts were not federally protected swaps that automatically shut state gambling regulators out. The election-contract issue was sent back down.

At the same time, the Third Circuit reached a different result earlier this year in New Jersey. Today, New Jersey asked the Supreme Court to take that case, putting the broader fight over state gambling law and federally regulated prediction markets in front of the justices for the first time.

Wisconsin has its own litigation moving too. A federal judge refused in July to block the state from enforcing its gambling laws, leaving its cases involving prediction-market platforms alive.

So an 1849 statute written in an era of tavern wagers is now sitting inside a modern fight involving prediction markets, federal commodities law, state gaming authority, a circuit split and the Supreme Court.

That’s exactly why Vulcan matters. The answer isn’t sitting in one statute or one case. You have to follow the old law into current agency guidance, active litigation, appellate decisions and whatever happens next at the Supreme Court to understand what actually governs today.

08/25/2026

Five federal agencies still have active rules about protecting typewriter ribbons.

The Federal Energy Regulatory Commission says an insecure ribbon can’t be left in the machine overnight. The White House Office of Management and Budget says ribbons should be removed after each use. Other federal rules still explain how classified ribbons should be stored or destroyed.

And the logic actually made sense. A used ribbon can preserve a readable imprint of everything typed through it, so in the typewriter era, leaving one behind could mean leaving a copy of classified work sitting on the desk.

The machine disappeared. The security instructions stayed.

Here’s the best part: the Commodity Futures Trading Commission amended one of these rules in 2024, and the language covering “carbon paper, typewriter ribbons, recording discs, spools and tapes” survived.

That’s how old law hangs around. A section gets opened to solve one problem, while language written for another technological era can keep operating beside the new text.

With Justinian, can see the rule as it exists today, how it got there, which pieces survived decades of amendments, and how it fits into current law.

There is apparently no janitor for the Code of Federal Regulations.

08/20/2026

Three months from now, GTA VI comes out. Millions of people will hit a button on the PlayStation Store that says **BUY**.

Sony’s own terms are very specific about what that word means.

Section 10.1 says words including “own,” “ownership,” “purchase,” “sale,” “sold,” “sell,” “rent” and “buy” do not mean ownership of the content has been transferred to you.

Section 8.4 goes even further: when you order from the PlayStation Store, “you buy a personal license to use that product.”

And Section 10.2 contains the word worth paying attention to: **revocable.**

PlayStation content is licensed on a “non-exclusive and revocable basis.”

That gets especially interesting as physical media disappears. Rockstar has said the physical edition of GTA VI will contain a download code inside the box. No disc. The box ultimately points you back to the digital license.

States have started paying attention to the gap between what the checkout button says and what the agreement actually gives you.

California’s AB 2426 restricts sellers from using words like “buy” or “purchase” for digital goods unless they disclose that the transaction is a license. Maryland followed, and similar proposals have appeared elsewhere.

This is exactly the kind of fine-print question Justinian can dig through across contracts, statutes and jurisdictions to see what those terms actually mean once the law around them is taken into account.

Read the terms. Then read the law around them.

Or let Justinian handle both.

08/18/2026

Congress passed a law describing an owl’s outfit.

16 U.S.C. § 580p still defines Woodsy Owl as “a fanciful owl, who wears slacks (forest green when colored), a belt (brown when colored), and a Robin Hood style hat (forest green when colored) with a feather (red when colored).”

Congress legislated the pants. The belt. The hat. And it wrote the colors as conditional because somebody in 1974 thought ahead to the black-and-white version.

It goes further than the definition. Unauthorized use of Woodsy Owl “may be enjoined... at the suit of the Attorney General.” Active Forest Service regulations also run a licensing regime on top of it, including a requirement that the Chief of the Forest Service find you “well qualified to further the goals and purposes of the Woodsy Owl campaign.”

Fifty-two years later, all of it is still current law.

That last part is where the research gets interesting. Finding an old, ridiculous statute is easy. Knowing whether it still binds means following it into the regulations built around it, checking what remains active, and making sure nothing later changed the answer.

That’s what Justinian did here.

This is somehow still a law.

08/12/2026

You pay for access to the legal database. That doesn’t mean your AI agent gets access too.

LexisNexis’s terms restrict third-party autonomous agents from logging in, navigating or querying the service without prior written permission.

And the restriction can reach human workflows too.

Lexis says use may become “systematic” when its service is functioning mainly as the data source for another system, even if a person is manually moving the information. So simply putting a human in the loop doesn’t necessarily solve the problem.

That creates a deeper issue for legal AI built on someone else’s database. If you can’t freely use that database as the data layer behind your system, the database owner can constrain how the product is built and how far its research workflows can go.

That’s why control of the corpus matters.

Vulcan built Justinian around a legal corpus it controls, allowing it to follow authority, move across sources and keep researching without another platform setting the limits.

07/23/2026

For 91 years, Humphrey’s Executor stood between the president and the independent agencies.

The 1935 decision limited the president’s ability to fire leaders of the FTC, FCC, SEC, and other independent agencies at will. It was cited 805 times and treated as bedrock law.

Then, on June 29, 2026, the Supreme Court tore it down.

In Trump v. Slaughter, a 6–3 majority overruled Humphrey’s Executor and held that the president can remove FTC commissioners without cause.

That creates a serious problem for anyone using general AI for legal answers.

Ask whether a president can fire an FTC commissioner and a model may still tell you no. It can cite Humphrey’s Executor with complete confidence because it learned from decades of legal text treating the case as settled law.

The answer sounds authoritative. The law changed 5 weeks ago.

Vulcan Technologies reads the live legal record and tracks how every case, citation, and authority connects. When a precedent falls, Vulcan sees the decision that killed it and the exact date the old rule stopped being good law.

07/20/2026

n Wisconsin, serving fake butter can cost you a year in jail.

Wis. Stat. § 97.18 says a restaurant cannot serve colored margarine as a butter substitute unless the customer asks for it. And a repeat violation can mean up to 1 year in county jail.

America’s dairy state put real criminal penalties behind butter protection, and the statute is still active.

Vulcan Technologies is built to search across that entire legal corpus, identify the law that actually applies, trace its history and connections, and return the source document behind the answer.

Because legal research only works when the system can find all of the law, including the part nobody expected to matter.

Vulcan Technologies

09/16/2026

09/14/2026

09/10/2026

09/02/2026

08/25/2026

08/20/2026

08/18/2026

08/12/2026

07/23/2026

07/20/2026

Address

Website

Alerts

Shortcuts

Share

Category

Listing takedown request