Good morning, and happy Monday.

As the weather gets warmer, your daily reminder that going for walks in the woods is super good for you.

Welcome to your week.

— Ian Krietzberg, Editor-in-Chief, The Deep View

In today’s newsletter:

• 🩺 AI for Good: Conformal classification

• 🏛️ Musk’s OpenAI lawsuit survives a push for dismissal  

• 🏛️ AI legal battles: Even for Meta, it’s all about fair use

AI for Good: Conformal classification

The application of machine learning and AI techniques in high-stakes environments, like healthcare, for instance, presents a massive challenge considering the persistent unreliability of the models and systems in question. 

Whether we’re talking about vision models or language models, we’re talking about predicted output; those predictions aren’t always accurate. 

There exists a technique, called conformal classification, that can, to a degree, remedy the uncertainty problem. Conformal classification replaces a single prediction with a range of probabilities that come with a guarantee that the correct probability — in this case, a diagnosis — is somewhere within that range. 

• Think of a doctor using a medical imaging model to analyze chest X-rays — that doctor would want to be sure to actively rule out lung cancer and any other possible diagnoses before accepting a single prediction as true. The challenge here is compounded by the fact that many conditions show up with remarkable similarities on medical imaging. 

• The problem with this technique is that, in order to ensure the correct probability gets included in the final output, models tend to output too many predictions for it to be useful. 

A team of MIT researchers recently applied a technique called Test Time Augmentation to conformal classification, a simple method that massively reduces the final output size without requiring researchers to retrain a given model. 

Why it matters: More accurate, more actionable sets of predictions — outputted with confidence scores — could make the technology more genuinely accessible and usable in the medical field. Importantly, it could also reduce some of the risk associated with implementing a system that is inherently unreliable. 

Maximize AI For Your Small Business

As a small business owner (or employee), you know the value of finding every little edge or advantage when it comes to getting things done – and AI just might be the greatest hack of all. But with the constant influx of new news, information, and tools, it can get overwhelming… which is why Salesforce has pulled together this free guide to help you out.

The Guide To AI For Small Businesses

In the guide, you’ll find everything you need to make the most of AI for your small business. Whether you’re looking for a leg up with the best strategies or simply want to see how other small businesses are putting AI to work - this guide has you covered.

Download your free copy from Salesforce right here.

Musk’s OpenAI lawsuit survives a push for dismissal 

Elon Musk’s lawsuit against OpenAI and Microsoft last week survived a dual attempt by both defendants for dismissal, according to a court filing. 

Musk’s litigation against the startup has been ongoing for more than a year. His latest goal is to prevent — or at least, slow down — OpenAI’s pending restructuring from a non-profit into a partially for-profit enterprise, a transition that has become the main qualifier for OpenAI’s last two multi-billion-dollar fundraising efforts. 

Judge Yvonne Gonzalez Rogers of the Northern District of California granted motions to dismiss outright eight of Musk’s allegations, but five of those motions were denied. 

• Musk’s claims of breach of contract were dismissed, but Gonzalez Rogers said that “Musk adequately pleads in the alternative that there is an implied-in-fact contract manifested by the alleged conduct of the OpenAI defendants.” As such, Musk’s claims of a breach of quasi contract and unjust enrichment survive. 

• Importantly, Musk’s allegation of fraud against Sam Altman, Greg Brockman and OpenAI survive, with the judge writing: “Musk adequately alleges that the defendants promised to maintain OpenAI’s non-profit status and structure in order to obtain his contributions, and that they intended to do so in order to obtain the capital needed to create a for-profit venture to enrich themselves.”

Prior to this, the judge denied Musk’s request to prevent OpenAI’s transition to a for-profit organization, promising instead to expedite a trial. That trial is currently scheduled for March of 2026, several months after OpenAI’s year-end deadline to complete the conversion or lose a massive slice of its latest funding round. 

Last month, OpenAI countersued, claiming that Musk’s legal harassment is a thinly-veiled attempt to “seek a competitive advantage.”

▶️ Live webinar: 5 Hours of AI Strategy for Marketing Agencies and Web Professionals

Join us for a 5-hour mega webinar packed with practical ways to use AI to work smarter and deliver better marketing results.

There’s a lot of hype out there—but this event cuts through the noise with real strategies, real tools, and real success stories.

• 9 AM PT – Boosting visibility with Bing’s Fabrice Canel Learn how to drive AI-powered clicks, manage crawlers more effectively, and improve rankings by signaling content updates.

• 10 AM PT – Scaling your marketing in 2025 with strategy and AI John Jantsch, founder of Duct Tape Marketing, shares how to streamline operations, attract better clients, and grow your business with smart strategy and the right tech.

• 11 AM PT – AI-powered agency success stories Agency leaders Nat Rosasco, Josh Neimark, and Ruan Marinho discuss how AI is reshaping their business models and delivering better results.

• 12 PM PT – Fast AI wins for agencies, marketers, and web pros Joseph S. Kahn and Ruan Marinho reveal simple strategies for spotting opportunities, crafting stronger offers, and increasing conversions.

• 1 PM PT – Hands-on prompting that works Andy Crestodina, Katie Robbert, and Santi Clarke will walk you through building your own reusable prompt library to scale content and marketing tasks with ease.

Register once, join when you can, and get the recordings sent straight to your inbox.

[Save My Seat]

• The rise of robotaxis: Uber last week announced a partnership with robotaxi firm Momenta, to introduce autonomous vehicles to the Uber platform. The move targets markets outside of the U.S. and China.

• Sinking NASA: The Trump Administration has floated plans to cut about $6 billion from NASA’s budget. A remaining $1 billion will be reallocated to focus on Mars missions, something that aligns with Elon Musk’s personal ambitions.

• Apple is working with Anthropic on an AI coding tool for Xcode (The Verge).

• Earnings show one tech segment starting to feel the tariff pinch fastest (CNBC).

• A DOGE recruiter is staffing a project to deploy AI agents across the US government (Ars Technica).

• DOGE put a college student in charge of using AI to rewrite regulations (Wired).

• Trump-linked stablecoin used in $2B Abu Dhabi Binance investment (Semafor).

AI legal battles: Even for Meta, it’s all about fair use

A few of the AI copyright lawsuits out there are what we might consider ‘big ticket items.’ 

The New York Times’ and Authors’ Guild suits against OpenAI are major heavy hitters. As is Kadrey V. Meta, a class action lawsuit of authors against Meta that alleges sweeping copyright infringement relevant to the construction of Meta’s generative AI models. An interesting element to this case, specifically, is its revelation — through internal communications shared with the court — that Meta knowingly trained its Llama models on datasets of pirated books. 

At a summary judgment hearing last week, Meta’s attorneys argued that “It would be impractical, and I don't think there's really any dispute about this, for a company like Meta to go to authors one by one for purposes of this training, in order to have an effective large language model.” 

But U.S. District Judge Vince Chhabria didn’t appear particularly moved by Meta’s arguments. Nor did he seem inclined to side with the plaintiffs, essentially expressing that both sides have a point. 

“You’re giving the product the ability to produce a million articles. You are dramatically changing — you might even say obliterating — the market for that person’s work, and you're saying that you don't even have to pay a license to that person for using their work to create the product that's destroying the market for their work,” Chhabria said, addressing Meta’s attorneys. “It’s not inconceivable to imagine that the market for the copyrighted works that were fed into the model could be eliminated entirely. How could that be fair use?”

• Court filings indicated in March that Meta does derive revenue from Llama, despite not selling access to the model in the same way that developers like OpenAI and Anthropic do. 

• More recently, Meta launched its Meta AI Assistant as a standalone app. 

These are important elements to the case; the fair use doctrine that developers have so often touted is not a one-size-fits-all framework. Applied on a case-by-case basis, the fair use argument must survive a number of pillars, such as the purpose of the use (commercial or something else), the amount of the copyrighted material used and the “effect of the use upon the potential market.” 

Meta’s use of the copyrighted material for commercial purposes, purposes that are very likely to severely impact the potential market, makes its claims of fair use challenging at best. This applies to the other developers who have touted fair use protection, as well. 

Meta’s attorneys said that the market impact Chhabria described is speculative, to which the judge responded: “I don’t know if you need to wait to have the impact. Potential effects can count.”

Chhabria said that if the plaintiffs could prove that Meta’s GenAI tools would severely impact how much they could earn from their work, Meta would be “destined to fail.” 

But Chhabria consistently pressed plaintiffs’ attorneys for evidence that the authors’ potential market would actually be impacted in such a way. “It seems like you’re asking me to speculate that the market for Sarah Silverman’s memoir will be affected,” he said. “It’s not obvious to me that is the case.”

Both sides had filed motions requesting partial summary judgment, which would allow certain claims to be cleared up before trial. In granting either side’s motion, Chhabria could establish a highly impactful precedent that could play into any of the other ongoing copyright cases. 

At the end of the hearing, Chhabria joked that he would issue a ruling that same day, then said: “Just kidding! I will take a lot longer to think about it.”

It’s not clear when that ruling will come, or what it will cover. But the plaintiffs, beyond damages, are seeking a permanent injunction that would force Meta to make changes to its models by removing the infringing data. 

There’s no way to do this after a model is trained — Meta would have to restart from scratch. 

And if the plaintiffs were to win — and if those victories are echoed in suits against other developers — it would surely open the floodgates for additional copyright infringement lawsuits until all the developers are forced to train only on content they have express permission to use, a costly, time-consuming endeavor that they desperately want to avoid. 

I wonder how all those media organizations that inked content licensing deals with OpenAI will feel if these lawsuits are actually successful. 

🤔 Your thought process:

Selected Image 1 (Left):

• “AI Store either has no entrance or it's blocked by a table. Talk about smart stores!”

Selected Image 2 (Right):

• “Shadows make sense.”

💭 A poll before you go

Thanks for reading today’s edition of The Deep View!

We’ll see you in the next one.

P.S. Enjoyed reading? Take The Deep View with you on the go! We’ve got exclusive, in-depth interviews for you on The Deep View: Conversations podcast every Tuesday morning. Subscribe here!

If you want to get in front of an audience of 450,000+ developers, business leaders and tech enthusiasts, get in touch with us here.