🧵 The historic NYT v.
@OpenAI
lawsuit filed this morning, as broken down by me, an IP and AI lawyer, general counsel, and longtime tech person and enthusiast.
Tl;dr - It's the best case yet alleging that generative AI is copyright infringement. Thread. 👇
2/ The visual evidence of copying in the complaint is stark. Copied text in red, new GPT words in black—a contrast designed to sway a jury. See Exhibit J here.
My take? OpenAI can't really defend this practice without some heavy changes to the instructions and a whole lot of
1/ First, the complaint clearly lays out the claim of copyright infringement, highlighting the 'access & substantial similarity' between NYT's articles and ChatGPT's outputs. Key fact: NYT is the single biggest proprietary data set in Common Crawl used to train GPT.
End/ The case could be a watershed moment for AI and copyright. A lot of folks saying OpenAI should have paid. We'll see!
What's at stake? The future of AI innovation and the protection of creative content. Stay tuned.
#OpenAI
#IPLaw
#Copyright
#NYTimes
🧵Elon's losing case against OpenAI, Microsoft, and Altman, as explained by me, a tech lawyer, general counsel and former litigator.
Tl;dr - PR fireworks and fun-to-read intrigue and philosophizing about AGI. But legally, a stinker because there’s no contract breach. Thread. 👇
3/ 🦸 NYT is a great plaintiff. It isn't just about articles; it's about originality and the creative process. Their investigative journalism, like an in-depth taxi lending exposé cited in the complaint, goes beyond mere labor—it’s creativity at its core.
But here's a twist:
@tomgara
What’s most fascinating about survivor to me (watched 25 seasons) is how similar it is to corporate America. Best education in interpersonal dynamics and power I’ve ever seen. Also notable - players who are substantively good (at challenges) rarely win.
5/ 😈 The complaint paints
@OpenAI
as profit-driven and closed. It contrasts this with the public good of journalism. This narrative could prove powerful in court, weighing the societal value of copyright against tech innovation. Notably, this balance of good v evil has been at
4/ ❌ Failed negotiations suggest damages for NYT. OpenAI's already licensed from other media outlets like Politico.
OAI's refusal to strike a deal with NYT (who says they reached out in April) may prove costly, especially as OpenAI profits grow and more and more examples
7/ 💼 Another interesting point: NYT got really good lawyers. Susman Godfrey has a great reputation and track record taking on tech. This isn't a quick cash grab like the lawsuits filed a week after ChatGPT; it's a strategic legal challenge.
6/ 🚫 Misinformation allegations add a clever twist. The complaint pulls in something people are scared of - hallucinations - and makes a case out of it, citing examples where elements of NYT articles were made up.
🍊 Most memorable example? Alleging Bing says the NYT published
🦜OpenAI seems to have fixed verbatim content parrot-backs, at least since NYT put together Exhibit J.
Some copyright-aware answers from ChatGPT ...
"I'm sorry, but I can't provide verbatim excerpts from copyrighted texts"
"I can't complete the paragraph"
"I can summarize or
2/ The visual evidence of copying in the complaint is stark. Copied text in red, new GPT words in black—a contrast designed to sway a jury. See Exhibit J here.
My take? OpenAI can't really defend this practice without some heavy changes to the instructions and a whole lot of
🧵 The state of legal AI, as explained by me, a 20-year lawyer, a three-time general counsel, and current legal AI company founder.
Tl;dr - the future is AI-enabled super-lawyers. It'll be here FAST.
H/t
@8teAPi
for inspiring me to write out what I've been seeing. 👇
Long discussion with senior partner at major Bay Area law firm today
> expects legal AI to decimate profession
> law firms charge by hour, and gen AI specifically cuts time for many many tasks
> unimpressed by most specific legal AI offerings
> chatGPT with some prompting is
OpenAI acting fast. No more Luigi.
"I'm unable to create an image with Luigi, as it doesn't align with the content policy for image generation. However, I can assist with a different request or modify the current scene in another way."
But lots of trademarked and copyrighted
🧵 OpenAI responded to the NYT’s lawsuit with a 1,063-word blog post yesterday.
Let's dig in. Thread from the X algo's new fave tech and IP lawyer and GC.👇
Tl;dr - the blog post is weak. Little data and odd citations. A missed opportunity for OpenAI, who has a good fair use
🧵 The historic NYT v.
@OpenAI
lawsuit filed this morning, as broken down by me, an IP and AI lawyer, general counsel, and longtime tech person and enthusiast.
Tl;dr - It's the best case yet alleging that generative AI is copyright infringement. Thread. 👇
@neylwalecki
I was waiting for somebody to find the prompt. Good work.
It’ll be interesting to see how the court handles prompting. AI terms of use say that you cannot tell it to do illegal things, which arguably intentionally prompting to misinform is doing. I think the misinformation part
"Copyright is not about the money. It's about ALL of the money."
@CeciliaZin
on the NYT lawsuit against OpenAI and Microsoft, which could reshape the industry.
👋 There's a meme. Every great FAANG engineer is resigning to join a web3 related startup. Such a meme even the NYT wrote about it. I'm not an engineer, but ...
I joined
@Replit
as GC & BD last month!
What
@amasad
and
@HayaOdeh
have built spoke to me. Why? 🧵 1/N
@paulg
Amazon stopped putting the actual products purchase in order confirmation emails many years ago for this reason (so Google couldn’t know them through Gmail).
Some personal news: after 2 amazing years, I’ve left
@Replit
! Incredibly grateful to
@amasad
and
@hayaodeh
for being the best founders out there and for their vision and incredible mission of empowering a billion software creators. In my time here, Replit:
📈 went from 10M to
@chi_urbanist
San Francisco's Italian consulate works an extra half hour than Chicago's lol.
Italian here, had to get a passport renewal appointment 22 months out with complicated online system. To be fair, they did send it the new passport later after my appointment. Bravi!
🚀 Luigi's revenge? More OpenAI progress on copyright today.
Ask for a copyrighted character. Now, DallE either refuses or -- critically for copyright -- abstracts your requests. It's a filter to avoid infringing outputs. Examples and legal analysis in thread. 👇
OpenAI acting fast. No more Luigi.
"I'm unable to create an image with Luigi, as it doesn't align with the content policy for image generation. However, I can assist with a different request or modify the current scene in another way."
But lots of trademarked and copyrighted
Tech lawyer here. I've counseled 100's of PMs across 4 co's, including FAANG. Nice 🧵on how product and legal teams work best as partners.
They get creative, co-develop, reach goals. Even w. high-lift laws like GDPR. Legal = advisor, not approver.
#lawtwitter
#inhousetwitter
A thread on working well with your partners on the Legal team:
(or, how to responsibly manage risk & relationships when conceiving & building innovative products)
1/ ✅ First, Musk filed in Superior Court in San Francisco, not federal court. Musk bases his case on state law contract claims.
He has to show a contract and that it was breached. Musk claims:
1️⃣ We agreed OpenAI and AGI would be open. But OpenAI is closed.
2️⃣ We agreed AGI
Lawyer friends - I launched a class on AI prompting for lawyers. Learning to prompt makes you (a) much faster and more effective, (b) avoid hallucinations, and (c) ahead of your peers. Even Chief Justice Roberts says AI has "great potential." Legal work "may soon be unimaginable
End/ 🎭 "The public is still in the dark", Musk says. He calls out "the irony and tragedy" of OpenAI becoming "Closed, For-Profit AI" and says he relied on Altman's (unwritten) word about it.
But there's no contract requiring transparency or non-profit.
So, Musk loses. End.
"Copyright is not about the money.
It's about ...
ALL of the money."
-overhead in a sleeper hit class in law school - a law & economics copyright seminar - from the guest speaker, a lawyer for a big rapper at the time, like P. Diddy.
4/ 📂 So what about the next line in the Articles, that OpenAI “will seek to open source technology for the public benefit when applicable.”
Psst - “when applicable” is magic legal code for “whenever we want.” It could be never.
Btw, OpenAI does open sources things. Even
3/ 🌎 Even assuming Musk were a party to OpenAI’s articles of incorporation - hint, he's not - let's look at what they say.
First, that the “technology will benefit the public” - got it.
OpenAI showing benefit from its tech is easy. Subjective. Impossible to breach.
✅
@LauraScottandCo
Love this, congratulations!!Lawyer with three girls (one I had in law school and another right after the bar exam) and a fourth on the way too. 💪🏼💕
6/ 🧠How about AGI? Why is that a thing here?
A stretch, but Musk is trying to show breach from Altman removing the old board members. Remember Helen Toner and Tasha Mccauley?
Musk says without them the board can't assess whether AGI benefits the public. Musk claims GPT-4 is
2/ The problem is no actual contract was breached. Let me explain.
Musk combines three things and says they're the "Founding Agreement" ...
1️⃣ Talks with Altman about what OpenAI was to be
2️⃣ A 2015 email to Altman where he says
3️⃣ OpenAI’s articles of incorporation, which
5/ 😈 What about Microsoft?
The complaint say OpenAI is "a closed-source de facto subsidiary” and "ostensibly non-profit golden goose" of Microsoft, a $3T company.
Nice rhetorical move. Meaningless legally.
Defendant's size isn’t a factor in a breach. Musk never had a deal
7/ 🌳 Then Musk's lawyer - Morgan Chu, known for patent trials - uses a fun analogy.
"Imagine donating to a non-profit whose asserted mission is to protect the Amazon rainforest, but then the non-profit creates a for-profit Amazonian logging company.... That's the story of
8/ ⚖️ Also notable: Musk picked SF over Delaware. Seems strategic, tapping into local techlash sentiments. It's an interesting move, considering Musk's, uh, mixed success in Delaware courts.
#LegalStrategy
@lessin
Disagree.
@Calendly
1 - enables 10x more "one shot, one kill" email interactions (h/t
@timferrisshow
); 2 - empowers the recipient to pick their fave time; 3 - saves everyone time; 4 - deals with time zones; 5 - creates the invite automatically. Lots of open windows help...
Text of the 2015 email - thanks
@TechEmails
Of note:
✅ "Good of the world" is in air quotes (!)
✅ The email talks about 'significant financial upside', which means Musk knew that's how it'd be set up
✅ The email says "we'd have an ongoing conversation" about open source. Aka
9/ 🌐 The rest of Musk's action? So many words for a lot of nothing of a case.
He underscores the societal issues surrounding the development and control of AGI, says he recruited people and helped Altman, and is worried about AGI.
@azeem
Based on what I know today, I think it’s a even jump ball between:
1 - litigating all the way to the Supreme Court. The New York Times has the money and is the perfect plaintiff, and Microsoft is very good at creating new law around technology. Like the betamax case in the 80s,
@jesslynnrose
A recruiter told me “you’re not really a tech lawyer” and “_maybe_ I can possibly get you an interview at [crappy place]”, then I got a job as counsel at Amazon a month later and became first lawyer on Alexa.
Meanwhile, OpenAI revenue up 20% in a month, to $1.6B ARR. Great to see the Altman affair didn't phase enterprise customers.
Great reporting by
@theinformation
.
@griswold
You're right that ChatGPT isn't great for contemporaneous news (yet - though Bing is). But I have continuously paid for access to both NYT and NYT Food archives for 10 years.
I access NYT recipes from the early 2010's about once a week (h/t Eleven Madison Park's amazing granola
🏡 In-house - at companies and not law firms - is the best place for legal AI.
We are relatively cheaper and don't bill by the hour. We get more done. We hire and fire firms. CEOs trust _us_.
As a result, in-house lawyers have grown 7.5x times the rate of other kinds of lawyers
@Replit
@amasad
@HayaOdeh
1 - Creators. Replit is a place to code, in any coding language with no setup, collaboratively. Today, Replit has over 10M+ users. Over 80% are outside the US. 40% women. 50% under 18. And this next gen of engineers? Looks more like the world! 🌏
I disagree the legal profession is decimated. The future I see is
🦸🏻♀️ Empowered super-lawyers, like the 100X AI-enabled software engineers
@amasad
and
@vkhosla
talk about.
🤝 Insider and niche knowledge will keep rates for certain expertise high. Sometimes you want the lawyer
@benoitraphael
Intriguing question. The complaint itself says the text was generated “with minimal prompting.”
The case is so new that the exhibits aren’t live on pacer (the federal courts’ system) yet. I’ll take a look later today and update. The relevant exhibit is exhibit J. 🙏🏻
@Suhail
Chiming in as the legal reviewer at companies of all sizes that scenario 1 ain’t all good, nor scenario 2 all bad. Scenario 2 at Amazon took only a few weeks ...
🤩 Tfw your new company does a kick-butt deal with your old company. A great series taught by the best, on a platform made for builders. ☕️ Java on
@Replit
offered by
@bloomtech
ftw.
Two months ago, Replit reinvented online learning with
#100DaysOfCode
.
Tens of thousands have signed up. It is the best python learning experience on the internet.
Today, we take it a step further.
Introducing 9 days of Java with
@bloomtech
@TonyW
@neylwalecki
The exhibit (linked below, h/t
@StewartMelissa0
) show the prompts were just article snippets as 2 sentences. ChatGPT just gave the rest.
From my testing now, it looks like OpenAI has already made changes.
I gave ChatGPT example 1 ... Its response - "I can't continue the text
@cristofrcharles
@OpenAI
Hm. Fair use requires showing the use is transformative - hard to show that here given the verbatim copies and that the responses are a substitute for paying for archives. Wdyt?
@jefielding
Interesting. Would be curious to hear the advice you gave him if you're up to share.
My take - having customers already is an advantage in building AI. For companies deep with customers and that offer something purpose-built, AI should be an accelerant.
Key point that as a general counsel I stress to those I advise. If you have to look at the text of a signed contract or board resolution when you’re in a dispute … it’s almost always too late.
OpenAI board is finding out there is a significant difference between de jure power and de facto power.
Board has de jure power, Sam has de facto power. Altman has hearts and minds, board has a piece of paper.
"Why Dora the Explorer Can't Come To Your Kid's Birthday Party ... The Issue Is Trademark Infringement"
Excellent 2008 piece by WSJ's
@katierosman
on legal actions from Marvel and others. They went after unauthorized birthday party entertainers and costume makers.
One
@hunterwalk
@gokulr
This article provides a great counter argument. Not having titles hurts women and URMs, particularly when they are more senior, because without a title as a signal people default to biases. And what about accountability? Speed? Clarity? Comp benchmarks?
@patio11
Lawyer version: a friend is a specialist litigator in ERISA (basically law on how co's manage retirement accounts). He had to stop going to continuing ed webinars because they were all teaching his cases.
Who's winning on the tech side?
🏆
@OpenAI
's GPT-4 is the only model good enough for legal work, and it's not even close. I say this having taught 1000s of lawyers how to use AI. But ChatGPT and vanilla tools have been instructed too friendly and tuned to be chatty and golden
PS - OpenAI's blog was about 1,000 words. NYT's complaint was 16,000 plus exhibits. Remind me who's the large language model again? 🫠
PPS - Of course ChatGPT helped streamline this post. It recommended the emoji for regurgitation: 💦. Despite my critiques, I really do love GPT
Me to then 3-year-old: how was your day?
Kid: Good! I made a project. How was your day, mom?
Me: I had a good day too. I did a project, and my work friend told me it was good.
Kid: Did you let it dry?
@jasonlk
Even as a former litigator, I strongly agree. Intellectual property disputes run 500K-1M a month to litigate on the very low end. Hard to imagine prepublic companies that can afford that extra burn.
This Masimo case is interesting though because it got national attention. My
"Mom, are you on the Supreme Court?"
So said my 7-year-old, when I told her SCOTUS was made up of the most important lawyers in the country.
She's a teen now. That dinner table conversation has inspired me to push harder in law and tech.
Happy
#InternationalWomensDay
!
@Replit
@amasad
@HayaOdeh
2 - Software at large - is still eating the world (ht
@pmarca
). Infinite economic opportunity goes along. Data: the percentage of assets of the S&P 500 that are intangible went from 17% in 1975 to 90% in 2020.
I have taught 100s of lawyers AI prompting with GPT, and my tests of Gemini revealed exactly this. Too many disclaimers in Gemini to be useful for legal work.
The difference between asking Google Gemini and Microsoft Copilot to write an ad for selling your goldfish.
Google has fundamentally lost its way because it feels the need to inject its politics into AI answers. Imagine if Google Search criticized legal things you searched for.
Then: DevRel for dozens through live hackathons in Sunnyvale with bad pizza
Now: DevRel is reaching 15M devs on
@Replit
with templates and people globally coding with them for fun
The Past: Devs in Silicon Valley code on $2500 machines
The Future: Millions build software together, everywhere, on Replit
We’re global, 15M users, mobile, collaborative, cloud-based.
Today, we take it further with verified company profiles. The new way to engage developers.
New feature announcement and privacy disclosure in the
@ChatGPTapp
just now.
"Your GPT can now learn from your chats"
📈Says GPT improves over time
⚙️ Adds a new "Reset memory" button in settings
🔒 New privacy disclosure in settings: "Your GPT will continuously improve."
You know when there’s new tech and big deals get done, and startups with the best UX and solution for users seem to come out of nowhere and win? That. Excited about this!
We're teaming up with
@googlecloud
.
Replit's 20M+ developers will get Google Cloud services, infrastructure, and foundation models. Idea to live software on Replit just got even faster.
@Replit
@amasad
@HayaOdeh
@pmarca
3 -
@naval
says it best: what would the world look like if _everyone_ could create software to solve any need? Replit, enables that.
The software field? Green. 🟩 [...]
The engineering team when they see that someone from legal has been typing non-stop for 18 minutes in the "Eng/Legal" Slack channel but hasn't hit send on the message yet
@Blanketman_01
Lawyer answer – it depends. I’m working on a thread on fair use here. It’s a four factor, squishy, some would say liberal artsy test. Notoriously difficult to predict.
Tired: Gary from Chicago asking Zuck about "meta days" off. RIP Gary.
Wired: Scrappy
@Replit
team goes from question in Slack to a 20-partner launch live to 15M users in five weeks.
People say we ship fast at
@Replit
... and we do!
We went from concept to launch w/ ~20 partners in <6 weeks. BD can move fast too!
Here are four photos showing how fast we move.