Subscribe
Share
Share
Embed
Everyday AI Made Simple – AI for Everyday Tasks is your friendly guide to getting useful, not vague, answers from AI. Each episode shows you exactly what to type—with plain-English, copy-ready prompts you can use for real life: budgeting and bill-balancing, meal and grocery planning, decluttering and home routines, travel planning, wellness tracking, email writing, and more.
You’ll learn the three essentials of great prompts (be specific, add context, assign a role) plus easy upgrades like formats, guardrails (tone, length, “no jargon”), and iterative follow-ups that turn “hmm” into “heck yes.” No tech-speak, no eye-glaze—just practical steps so you feel confident and in control.
If you’re AI-curious, and short on time, this show hands you the exact words to use—so you can save your brain for the good stuff. New episodes keep it short, actionable, and judgment-free. Think: your smartest friend, but with prompts.
Blog: https://everydayaimadesimple.ai/blog
Free custom GPTs: https://everydayaimadesimple.ai
Some research and production steps may use AI tools. All content is reviewed and approved by humans before publishing.
00:00:00
Welcome to the debate. So imagine you are investigating a highly sophisticated bank heist.
00:00:06
Okay, a bank heist.
00:00:08
Right, and during the course of this investigation you stumble across a detailed blueprint, but it's written by the bank's own security team. It maps out exactly how to bypass the alarms, you know how to crack thevault. Wow. Yeah, And it even includes like diary entries from the architects debating the moral bankruptcy of their own plan. Today, we are looking at a situation very much like that in the tech world. We're examining the legal fortress of Open AI's fair use defense in the massive multi district copyright litigation happening right now in New York, which is uh known as M D L thirty one forty three.
00:00:42
And specifically, I mean, we are diving into a fascinating collision of two completely different legal universes. Because, on one side, you have the explosive evidentiary record that was just unearthed in the California state law trial of Musk versus Altman, which is largely about corporate governance, right? Fiduciary duties.
00:01:02
Exactly. State law stuff.
00:01:04
Right. But on the other side you have federal copyright law in New York. So, the central question we are exploring today is whether the internal evidence of OpenAI's commercial motives, uh their executive behavior and structural admissions, From that California trial, completely dismantles their fair use defense in the New York copyright case.
00:01:25
And to anchor us, Our entire analysis today stems from a pivotal new brief filed on May twenty fifth, twenty twenty six. Magistrate Judge Ona T Wong just ordered a massive cross production of evidence.
00:01:40
A huge order.
00:01:41
Oh massive. She is essentially taking the sensational, frankly damaging transcripts from the Musk trial, And injecting them straight into the copyright litigation. Now, for the sake of our discussion today, I am approaching this from the trenches of high stakes civil litigation. I am looking at it through the eyes of a cynical class action attorney. Fair enough. And my position is that this newly produced evidence provides unprecedented litigation leverage. It proves bad faith, massive commercial intent.
00:02:13
And verifiable market harm that will systematically destroy OpenAI's fair, use defense before a jury.
00:02:19
Okay, well, And I am analyzing this through the strict doctrinal lens of a neutral copyright law professor. My position is that while the optics from the Musk trial are undeniably terrible for OpenAI. Terrible, yeah. They are bad, but state law fiduciary duty breaches do not seamlessly overwrite federal copyright doctrine. If you look at the strict statutory interpretation of the Copyright Act, specifically uh seventeen U S C section one oh seven, Open A I's transformational use defense remains legally viable, regardless of how messy their boardroom politics got.
00:02:57
Well, If you are listening to this and wondering you know why a California state trial about a nonprofit board matters in a New York federal copyright case, you aren't alone. But when I look at Magistrate Judge Wang's May twenty first order, which converts that Musk trial record into literal ammunition for the copyright plaintiffs, I see a death knell for OpenAI. A death knell. That's strong. I mean it. They have a massive summary judgment motion coming up on August fourteenth, where they want a judge to just toss this case out. But fair use is fundamentally an equitable doctrine.
00:03:37
It is rooted in fairness,
00:03:39
Right? Equity.
00:03:40
It requires the defendant to have operated in good faith. In the real world of litigation, When you have a founder like Greg Brockman writing in his November twenty seventeen journal, about how to reach a billion dollars while actively calling their own nonprofit maneuvers, morally bankrupt, I mean the defense collapses.
00:04:03
See, I completely disagree that it collapses. Because I think you are letting the sensational nature of a trial transcript blur the actual statutory lines. The Musk trial evaluated commercial purpose purely as a breach of California charitable trust obligations to a donor. But it proves intent. They were asking, well, they were asking, did you betray Elon Musk's donations? Under federal copyright law, Commercial intent is only one fraction of the very first factor of the fair use test.
00:04:30
A very important faction though.
00:04:32
But look at recent twenty twenty five precedents, Specifically, cases like Bart's v Anthropic and Kadri v Meta. Courts have consistently held that if the act of training, an A I model is deemed highly transformative, then the personal equity stakes of executives simply do not equal copyright infringement.
00:04:51
I see why you think that, but let me give you a different perspective. You cannot surgically separate the objective mechanics of an algorithm from the intent of the people who built it. Not under factor one of fair use, Which explicitly asks whether the use is of a commercial nature or for nonprofit educational purposes. Okay, But for years, OpenAI's public posture was that their massive data scraping was just for research. But the scale of the admissions from this cross production is staggering. We have Satya Nadella, the C E O of Microsoft, testifying on cross examination that there was a quote.
00:05:26
Clear commercial element from the outset.
00:05:28
I mean, no one is denying they wanted to make money eventually.
00:05:32
That is a direct admission against interest from their primary financial partner.
00:05:36
I don't deny that Nadella said that or that it looks awful on a slide deck, but does Nadella's understanding of a commercial element legally negate transformativeness? Because the Supreme Court has weighed in on this exact tension.
00:05:49
It absolutely shatters their educational research narrative. Nadella confirmed under oath nine point five billion dollars in recognized revenue. We have his April twenty twenty two email. This is seven months before Chat GPT even launched, Stating he didn't want Microsoft to be the next I B M,
00:06:06
Right? A competitive tech market.
00:06:08
That demonstrates a clear aggressive motive to monopolize intellectual property, and it predates any plausible research framing of G P T four. Furthermore. How on earth can a company claim a non profit educational purpose when their former chief scientist, Ilya Sutskever, holds a seven billion dollar equity stake? And, he wrote a fifty two page memo warning. The board that rapid commercialization was overriding their safety protocols.
00:06:37
Look, That is a wildly compelling argument for a jury that wants to punish corporate greed. I grant you that. But have you considered the Supreme Court precedents of Campbell v. Acuff Rose or the recent Andy Warhol case?
00:06:49
Sure, the parody case.
00:06:51
Right, in copyright law, subjective financial motive does not erase objective transformative use. In Campbell, the rap group 2 Live Crew made a highly commercial, purely profit driven parody of the song Oh Pretty Woman. The court said it was still fair use because the work itself transformed the original.
00:07:10
But that was a parody. This is a machine churning out data.
00:07:14
The principle holds though. The fact that Greg Brockman's journal literally asks, financially, what will take me to one billion dollars is provocative, yes, But it doesn't change the technical nature of what the algorithm actually does with the scraped data.
00:07:28
I think it does change how a judge views it.
00:07:30
Section one oh, seven of the Copyright Act does not say fair use applies unless the defendant is greedy. It looks at the purpose and character of the use, If training a large language model creates entirely new functional capabilities, the commercial wrapper around that technology is secondary.
00:07:46
I'm just not convinced by that line of reasoning because you are treating fair use like a sterile math equation, rather than a defense built on equity and fairness. To go back to my opening thought, You are looking at the blueprint of a bank heist and arguing that because the getaway car they built is highly innovative, the theft of the money should be pardoned.
00:08:06
That is that is stretching the analogy.
00:08:09
The Supreme Court clearly established in older cases like Harper and Row that a defendant's bad faith weighs heavily against a finding of fair use. Brockman wrote in his journal, and I quote, If three months later we're doing B Corp, then it was a lie.
00:08:22
Yeah, the optics are brutal.
00:08:24
He admitted it would be morally bankrupt to steal the nonprofit from Elon Musk. This isn't just about making money. It's about deliberately exploiting a charitable status to acquire vast amounts of training data. And public goodwill under false pretenses and then monetizing it. That fundamental deception corrupts the equitable balance a judge has to weigh.
00:08:46
I understand the litigator's desire to paint a picture of a grand heist, it plays incredibly well in a courtroom, but as a copyright purist, I have to push back on the parameters of that analogy. Copyright law isn't a physical bank heist. Data isn't a physicalvault that you empty out, leaving the original owner with absolutely nothing.
00:09:05
Try telling the New York Times they have nothing.
00:09:07
But intent only gets you so far. Even if a judge concedes that Open A I's intent was overwhelmingly, ruthlessly commercial, We still have to evaluate the nature of the use itself and its actual market impact. You can be the greediest executive in the world, But if your product doesn't actually copy or replace the plaintiff's work in the real market, there is no infringement.
00:09:29
Okay, so we're looking at factor three and four now. Exactly.
00:09:32
Which is why we have to look at the actual output of the model. Bringing us to factor three, the amount and substantiality of the portion used, and factor four, the effect on the potential market.
00:09:43
And that transition is exactly where the copyright plaintiffs now have the ultimate smoking gun: Project Giraffe. This is perhaps the most devastating piece of technical evidence intersecting with the Musk trial.
00:09:57
Explain Project Giraffe for anyone who hasn't read The Brief. Sure.
00:10:00
Project Giraffe is OpenAI's internal code named engineering program. It was explicitly built to implement guardrails against their models, regurgitating verbatim copyrighted training content. From a litigation standpoint, let me explain why this is a total nightmare for the defense. Let's hear it. If a defendant has to create an entire secret engineering division to stop their product from spitting out verbatim copyrighted text, It is an implicit institutional concession that verbatim reproduction is a core feature of the underlying product. It goes straight to factor three. It proves they took the whole expression, not just the underlying ideas.
00:10:42
I come at it from a completely different way, and I suspect OpenAI's appellate lawyers will too. Implementing guardrails can and legally should be framed as responsible deployment. Responsible? Yes.
00:10:53
The U S Copyright Office's own May twenty twenty five A I report, specifically addressed this dynamic. Attempting to filter out infringement in the output phase doesn't inherently prove that the ingestion phase, you know, the actual training was illegal. It simply proves the company is taking active steps to prevent copyright infringement when the user interacts with the model.
00:11:12
I don't think a jury will see it that way.
00:11:14
But think about the policy implications. If we legally punish a company for building safety guardrails, we create a perverse incentive. We would be telling A I developers that they are legally safe for doing absolutely nothing to protect authors.
00:11:27
That is an interesting point, though I would frame it very differently. Your argument is like a car manufacturer, arguing that installing airbags after a series of fatal crashes proves their initial design was perfectly safe.
00:11:41
Well, no, it proves they are fixing a known issue.
00:11:44
They built Project Giraffe guardrails precisely because the underlying mechanical design is meant to infringe on the exact markets that the plaintiffs occupy. Whether that is news summarization for the New York Times or code generation for software developers.
00:11:57
But the guardrails exist to prevent that overlap,
00:12:00
And we aren't just relying on the existence of Project Giraffe as a concept. Magistrate Wang ordered the production of seventy eight million and ten million Chat G P T log reservoirs on March tenth. Plaintiffs finally have the empirical data. We don't have to guess if the models substitute for the original works, We have tens of millions of logged interactions to prove direct market substitution under factor four.
00:12:20
I mean, The seventy eight million output logs are certainly the largest data set ever produced in an AI fair use case. I'll give you that, but let's remember what courts actually look for under factor four. Factor four asks if the secondary use actively supplants the market for the original work.
00:12:39
Which it does.
00:12:40
Look at the Kadri v Meta decision. That case turned largely on a failure of proof regarding market harm. Meta's primary product, a social network, Wasn't commercially substituting for the plaintiffs books. In this M D L, The plaintiffs have to prove that Chat G P T users are using the model instead of buying a subscription to The New, York Times or buying a specific author's novel.
00:13:03
The logs will show that.
00:13:05
Open A I will argue that verbatim regurgitation in those logs is rare. It's idiosyncratic. It's usually the result of a user prompted edge case. They're essentially arguing that the user is acting as the infringer.
00:13:20
And How can OpenAI honestly maintain that narrative of, you know, rare idiosyncratic edge cases when Vinny Monaco shows up to testify?
00:13:30
Ah, yes. The Monaco deposition.
00:13:34
Right. Monaco was OpenAI's thirty B six corporate designee. For those unfamiliar, a thirty B six witness isn't just some guy from the I T department. He is legally designated to speak as the avatar of the entire corporation. When he speaks, OpenAI speaks.
00:13:52
And he did not speak very well.
00:13:54
He showed up completely unprepared for his deposition on Project Giraffe. Magistrate Wang explicitly found that his pattern of objections, quote, impeded, delayed and frustrated the fair examination of OpenAI. She ordered him back into the chair for three and a half more hours.
00:14:12
Which is humiliating.
00:14:13
In a civil trial, when your corporate avatar hides evidence about your anti- plagiarism guardrails, that is the textbook predicate for a Rule thirty seven adverse inference instruction.
00:14:24
That is a that's a severe penalty. A judge essentially tells the jury, The defendant hid this evidence, so you are legally allowed to assume the evidence would have been highly damaging to them.
00:14:37
Exactly. It is a litigator's worst nightmare. A jury will be told they can infer that whatever Monaco was hiding about Project Giraffe proves the model is essentially a piracy machine. Combine that adverse inference with the ninety two billion dollar projected return on investment from the Brad Smith memo, and the factor four market substitution argument practically writes itself.
00:15:01
I will certainly concede the optics of a botched thirty b. Six deposition are uniquely terrible, but optics don't override the statute. We have to pivot away from what a jury might think and look at the procedural realities of how this evidence actually enters the record. Okay. Because, the strongest evidence in the world is utterly useless, if the judge won't let the jury see it in the first place. You are relying heavily on using the Musk trial transcripts as impeachment material, specifically targeting Sam Altman's character to defeat summary judgment.
00:15:35
Absolutely, I am. The impeachment value here is astronomical. In this, it structurally breaks open AI's case. Summary judgment requires there to be no disputed material facts, right? Correct. Sam Altman testified in the California trial on May thirteenth, twenty twenty six under oath. I believe I am an honest and trustworthy business person. Yet in that same trial we have four contemporaneous insiders: Ilya Sutskever, Mira Murati, Helen Toner and Tasha McCauley all testifying under oath.
00:16:07
That Altman is deceptive or fosters a culture of lying.
00:16:10
It was a brutal week in court.
00:16:12
Moradi literally testified that Altman lied about a safety review, having been cleared by outside counsel.
00:16:18
Right, but how does skate court boardroom drama translate to a federal copyright claim?
00:16:23
Because OpenAI relies heavily on Altman's sworn declarations to characterize their AI training as a transformative public benefit under the Federal Rules of Evidence, specifically Rule six eight b, Which allows you to attack a witness's character for truthfulness. This material completely destroys their narrative. So you impeach him. Exactly. If the plaintiffs can prove the C E O is fostering a culture of deceit, his sworn statements about educational intent become highly disputed facts. The judge can't grant summary judgment, he has to send it to a jury. The equitable defense of fair use evaporates when your primary witness is impeached by his own executive team.
00:17:06
I urge extreme caution with that strategy, though. You are assuming Judge Sidney H. Stein is going to allow a massive detour into state law character assassination in the middle of a highly technical, multi- district copyright case. We must look at the structural risks of the M D L.
00:17:25
But Wang ordered it.
00:17:26
Magistrate Wang ordered this cross- production, but a magistrate does not have the final word. Judge Stein is the supervising federal judge. And, we already have a very clear signal of how he views these sprawling discovery disputes. Just earlier this year, Judge Stein fully reversed Magistrate Wang on the Books One and Books Two privilege fight.
00:17:47
Ah, the Lisa Blatt appeal.
00:17:50
Precisely, Lisa Blatt argued that appeal for OpenAI, and Judge Stein recloaked OpenAI's internal Slack messages about pirated training sets, finding they did not waive attorney client privilege. Stein is reading OpenAI's evidentiary objections incredibly seriously.
00:18:06
Sure, he's cautious.
00:18:08
He is willing to overrule his magistrate to protect the boundaries of the case. If you try to drag in days of testimony about California corporate board drama, Judge Stein is highly likely to exclude it under Rule 403. That rule basically says a judge can block evidence if its probative value is substantially outweighed by the danger of unfair prejudice or confusing the issues for the jury.
00:18:28
I mean, I acknowledge the Rule 403 risk.
00:18:31
Stein could easily say, you know, Altman's relationship with Helen Toner has nothing to do with whether a neural network infringes on a copyright.
00:18:38
A judge worried about a circus will shut down the clowns. And Stein's reversal on the books. One slack messages was a blow to the plaintiffs, definitely. But look at how brilliantly Magistrate Wang specifically drafted her May twenty first cross production order. How so? She anticipated Judge Stein's strictness. She didn't just grant the Musk transcripts generally, She explicitly framed the cross production around Open A I's commercialization efforts, Stating in her order that they are quote pertinent to the initial factor of the fair use analysis. She preemptively locked in the relevance to factor one.
00:19:14
That is a clever move.
00:19:16
Bytying the Musk evidence directly to the statutory language of section one. Oh, seven, She made it very difficult for Stein to throw it out on a rule, seventy two, a objection. Without looking like he is ignoring the very text of the Copyright Act.
00:19:29
Yes, Wang framed it that way, but framing doesn't bind the supervising judge at summary judgment. Open A I will undoubtedly cite cases like Thompson Reuters v Ross, arguing that even among direct commercial competitors, the legal focus must remain on whether the technological use itself is transformative. They'll try and they will point out the fundamental mismatch here. The Musk plaintiffs framed commercial purpose as a breach of obligation, Running to a specific donor. The M D L plaintiffs are trying to shoehorn that exact same breach into a copyright case. It's the exact same factual record, sure, but the doctrinal hooks are completely different. The law simply does not always allow you to easily transplant a fiduciary breach into a copyright infringement.
00:20:11
I disagree that it's a transplant. I really do. It's an unmasking. The public posture Open A I maintained from twenty seventeen to twenty nineteen was a deep commitment to a nonprofit structure. Citing broadly distributed benefits to humanity. That was their shield,
00:20:27
Which they leaned on heavily. Right, But the reality exposed by the Musk trial is Greg Brockman writing conclusion is we truly want the for profit. The public posture of Microsoft in this copyright MDL is that they are just an arms length investor. And the data training was OpenAI's independent research.
00:20:48
And what did Nadella's email expose?
00:20:50
The reality, Exposed by Satya Nadella's April twenty twenty two email, Is Microsoft aggressively taking a massive dependency on Open A I to exert genuine influence at every layer of the stack. We are taking Open A I's carefully crafted, public copyright defenses and impeaching them with their own internal contemporaneous corporate admissions. And what happens when Open A I's legal team successfully argues that their public posture regarding their corporate charter. Has zero bearing on the physics of their product. What if they prove that G P T four's ingestion of a New York Times article genuinely transforms factual data into statistical weights, regardless of what Greg Brockman wrote in his diary?
00:21:33
Well, If they can prove that.
00:21:35
Copyright law is inherently insulated by these strict transformativeness requirements for a reason. If, we allow a C E O's personal equity stake or an executive's lies to a board member. To dictate copyright liability, we jeopardize the entire software industry's ability to innovate. We would be saying that a nonprofit can legally train an AI because their hearts are pure, but a for profit doing the exact same computational process is a pirate. That is an incoherent standard for intellectual property.
00:22:02
I hear you, but it is only incoherent if you ignore the actual market harm, which brings us full circle. A nonprofit doing purely academic research, doesn't project nine point five billion dollars in revenue. By actively substituting the market for investigative journalism in software coding.
00:22:20
That is a fair point.
00:22:22
The combination of Nadella's admissions, Brockman's journal, and the technical concessions of Project Giraffe provides an overdetermined, inescapable record. We are moving into the August twenty twenty six summary judgment cycle with a record of commercial exploitation that is simply fatal to their defense. Well,
00:22:40
While I agree the optics are undeniably challenging. And the sheer volume of the factual record is unprecedented. I maintain that conflating corporate greed with copyright infringement is a fundamental error in logic. A judge focused on the mechanics of the law, like Judge Stein, Will prioritize the strict technical definition of transformative use over the emotional, equitable bad faith narrative you are building.
00:23:05
Well, We both clearly agree on one thing: The evidentiary record moving into this summary. Judgment phase is the most developed. Anti fair use record ever assembled against an AI company. No prior group of copyright plaintiffs has ever had this level of access to the internal machinery and motives of a defendant.
00:23:24
Agreed, it is a monumental convergence of state and federal litigation, And there are still several looming variables that could tilt the scales in either direction. We are still waiting on the Second Circuit's decision in the Raw Story case regarding D M C A claims.
00:23:40
Oh right, the Article three standing issue.
00:23:43
Exactly. Basically whether the plaintiffs have the legal right to sue over the removal of copyright management information, even if they can't prove direct financial injury. A reversal there could dramatically reopen a whole new front of D M C A exposure for Open A I and the M D L.
00:24:00
And, let's not forget what might surface when Vinny Monaco is finally forced back into the deposition chair for his court, ordered three and a half hours on Project Giraffe. If he has to produce the underlying engineering documents, you know, the guardrail design memos and the actual regurgitation rate metrics.
00:24:16
If they have those metrics.
00:24:18
Right. If those metrics confirm what we see in the seventy eight million logs, the mechanical arguments for factor three and factor four become mathematically undeniable.
00:24:28
It will certainly be a critical deposition. Ultimately, this M D L will force the judiciary to make a profound choice. They have to decide whether the future of copyright in the age of A. I should be dictated by the subjective intent and corporate maneuvers of executives or by the objective algorithmic mechanics of the technology itself.
00:24:49
And that is the billion dollar question. When you look closely at this legal fortress Open A I has built, you have to ask yourself, does it matter how brilliantly engineered the technical walls are? If the architects themselves documented exactly how they stole the land it sits on,
00:25:07
A question that Judge Stein will soon have to answer.
00:25:10
Exactly. Thank you for joining us on the debate. Keep questioning the architecture. We'll see you next time.