Brad Smith

"Only about 3% of consumers pay for AI services, spending about $12 billion, according to a recent study conducted by Menlo Ventures. The firm’s survey of 5,000 AI users concluded that people use chatbots and other models for a variety of tasks that is so diffuse, “there’s still a long way to go to everyday adoption.” "While most executives, investors and researchers at the vanguard of AI development see the technology eventually disrupting virtually every industry in the world, many have begun to acknowledge that making money from it will be a far more vexing challenge." The one lesson I learned from the dot com bubble was that when consumers are not paying the full cost for the services they are consuming - bad things are sure to happen. In that bubble, most of the subsidy was provided by venture funding. Here, it is venture funding + the war chests of Big Tech. At the end of the day, the subsidies have to end. Does that mean we AI is doomed - far from it. Just like in the wake of the dot com bubble, there are likely to be winners and losers, although some of the winners will limp along for a while until the market catches up with them.

6

Great insights from the IP Roundtable at Loeb & Loeb's AI Summit! Key themes included the importance of retaining ownership of AI-generated outputs and the need for robust legal protections against content scraping. A must-read for anyone navigating the evolving IP landscape. #IntellectualProperty #AISummit #LegalTrends

1

The Women in eDiscovery, San Diego are hosting a conference on October 18th in Irvine, CA to discuss the impact and opportunity that AI presents for e-discovery professionals and litigators.

7

PSA: The best conversations don't always happen on stages... sometimes, they start with sticky notes! 📝 Last week, I had the opportunity to lead a roundtable discussion with co-host, Benjamin Alred, at the inaugural WIPR - World IP Review AI & IP Summit in NYC. The topic was a big one: “Leveraging AI to maximize the power of IP as a cross-functional asset in your organization”. To make sure the conversation didn't get dominated by a few loud voices (always a risk!), we borrowed an idea I loved from The LegalTech Fund's Corporate Legal Retreat earlier this year. Everyone at the table rapid-fire jotted down their thoughts independently, first on how AI is currently being used in their IP processes; then on how they imagine their organizations and the IP ecosystem functioning wayyyyy into the future - just kidding, we targeted 10-15 years from now. It's all relative, right! The present was exactly what you might expect! A flurry of great tools, mostly point solutions. A bit of connectivity here and there (often IP and R&D, or IP and the C-suite), but the relationships were shallow and mostly one-directional. The future, though, looked radically different. The room imagined seamless workflows spanning ideation to monetization, heavy automation, and even some speculation about patent office processes running with little to no human involvement. Whether that future is desirable from the group's vantage point... the jury is still out on that one! 😅 The exercise was simple but the contrast was striking. Today, we have an ecosystem that’s diverse yet fragmented. Tomorrow, we might have one that’s wholly interconnected, automated, and operating at a completely different scale. The point is, we’re still just scratching the surface of what's possible. And if this conversation is any indication, we've got a brave new world ahead!!

76

What does it take to scale AI today? MoFo’s Justin Haan will speak on the panel, “AI Infrastructure: Building Stargate & Beyond” during SXSW on March 17. Justin will be joined by OpenAI's Head of Infrastructure Business Development, Matthew Castle, VP Strategy and Operations – Infrastructure, Peter Hoeschele, and Compute Delivery Lead, Kaylen Bushell, CRL for a discussion that explores how data center development has become mission-critical, how decisions about where to build are made, and what happens when AI growth collides with real-world limits. More information can be found here: https://gag.gl/odsbLh

13

California has set a new benchmark for AI oversight. Governor Newsom has signed SB 53, the Transparency in Frontier Artificial Intelligence Act (TFAIA), the first law in the nation focused on safety and accountability for “frontier” AI models. Read highlights of the TFAIA and explore what this means for the broader tech landscape. https://gag.gl/agJvVd

3

A fascinating development has been brewing in the privacy flank of the United States. On September 29, 2025, California Governor Gavin Newsom has signed the Transparency in Frontier Artificial Intelligence Act (SB 53) into law. This legislation establishes one of the most comprehensive regulatory frameworks for artificial intelligence in the U.S. privacy landscape. Notably, it represents a scaled-down yet more focused approach compared to the more stringent SB 1047 that Newsom vetoed in 2024. Interestingly, the primary emphasis of SB 53 seems to be localized on transparency rather than liability. Now, SB 53 introduces certain rarely-seen compliance requirements for the AI industry. It establishes the US's first transparency requirements for safety plans on the most advanced AI models - hence, the largest AI companies are mandated to publicly disclose their safety and security protocols in redacted form to safeguard intellectual property. Furthermore, companies must report safety incidents of critical severity within 15 days to the Governor’s Office of Emergency Services. These incidents include model-enabled CBRN (chemical, biological, radiological, and nuclear) threats, major cyber-attacks, or loss of model control. One of such rare aspects of SB 53 is its introduction of requirements for companies to disclose safety incidents involving dangerous deceptive behaviour by autonomous AI systems - the very first of its kind globally. This includes instances where AI systems misrepresent the effectiveness of their safety controls during testing, significantly increasing the risk of catastrophic harm. Unlike the EU AI Act, which mandates private disclosure to government agencies, SB 53 mandates public disclosure of safety protocols - with the purported aim of ensuring greater accountability and transparency. Additionally, SB 53 establishes “CalCompute” - a public cloud compute cluster that provides AI infrastructure for startups and researchers. This aim - to democratise access to critical computational resources and advance an industrial policy for AI. Furthermore, the legislation provides for enhanced protections for whistleblowers who come forward with evidence of critical risks or violations at AI companies. The State Attorney General’s office will enforce the law through civil penalties, although it doesn’t impose any new liability for harms caused by AI systems. What catches attention, something really interesting about the legislation, is that it has garnered support from a leading AI company (Anthropic). Rare as this singular support may be, for obvious reasons the move has also been facing flak from other major tech companies and industry groups. The link to the text to SB 53 - https://lnkd.in/dKPdBMEx Peritum Partners #AITransparency #SB53 #FrontierAI #CaliforniaAI #AISafety #TechPolicy #AIRegulation #ResponbleAI

6

LLMs are powerful — but raw models aren’t built for legal work. Here’s what they can (and can’t) do, and how Legal AI makes them safe for high-stakes environments.

30

2 Comments