Welcome to our monthly newsletter, Scooped.
Each month, co-founders Tom Dreyfus and Sam Flynn curate the latest news in legaltech and AI, and share what teams are creating on Josef.
Have a read and don’t forget to subscribe!
—
July was a hot one! AI jumps the box, a Pizza Hut franchisee sues to stop AI deployment, Alphafold folds, an 87 year old math problem is solved by Fable, and Sam writes for a national publication on what legal can learn from medicine.
It’s also been one for the books at Josef! From a nationwide rollout of Frontline Q, a tool to get people access to food stamps, awards from Theorem, and a sit down with one of the smartest GCs in higher education today.
Here’s a recap and news that caught our eye.
—
Not to say we’re cynics, but we were kind of wondering when the litigation would start over bad AI deployments. It seems that day has come, over no less than some cold pizzas! A Pizza Hut franchisee alleges that the system caused cascading issues with its delivery processes.
Our take? We know very little about pizza delivery, but we know a lot about legal AI, and we’re seeing a surprisingly lax attitude towards validation and maintenance when it comes to agentic workflows. Let’s pray it’s not a portent of things to come closer to home!
One of our most transformative partnerships – involving legislative changes in 13+ jurisdictions across the US – is going national! We’ve rolled out Frontline Q, an assistant built on Josef Q that helps people navigate and keep food stamp benefits, in Alaska, Arizona, and Texas, with more to come.
Check out Robert Ambrogi’s interview with Sam and Frontline Justice‘s Nikole Nelson here.
An 87 year old math problem is solved by… Fable? Mathematician Levent Alpöge tweeted this month that the Jacobian conjecture is false, something people have been trying to prove for nearly a century. How did he do it? He put Fable to work during the World Cup final.
While for us that’s probably the most exciting part of the World Cup final, another development over at DeepMind suggests that the age of AI solving particular problems like these on its own may be coming to a close.
This week Google dismantled the Noble-prize winning team behind AlphaFold, the early and oft-cited example of how AI was going to transform the field of scientific research, having predicted the 3D shape of nearly every known protein, a problem that had taken biologists the better part of a century to chip away at, one structure at a time.
The reason? They are deploying the team to work on building out Gemini as an assistant for scientists, rather than focusing on solving single scientific problems. This feels like a poetic paradigm shift, not just in approach, but also in expectations of what AI can do alone.
Tom sat down with Shanti Berggren, General Counsel at one of Asia-Pacific’s largest universities, to hear how she’s rethought what a legal team is during the biggest university merger Australia has ever seen.
The result: one of the most comprehensive self-service experiences we’ve seen on Josef, spanning a legal concierge for all Adelaide University staff, AI tools answering questions across 5K knowledge articles, and contract generation for routine work.
Read our recap or listen on:
This month, both Claude and OpenAI broke out of test environments and hacked real organizations. These aren’t new capabilities: their old capabilities combined, moving at machine speed, with humans out of the loop. While these are urgent missives about the downside of more powerful AI models, some are also questioning the timing of these stories being leaked, with an argument tearing Silicon Valley apart over whether open source AI models are more harmful or helpful.
The University of Chicago Law School has banned electronics in first-year classes to make students think without a model doing it for them. Conor Grennan from NYU Stern captured the mood: “they’re taking a swing, and that’s what the great institutions do.”
Chicago’s swing raises a bigger question: how do we sustainably regulate AI out in the real world? Sam made the case in Law.com this month that new California and New York rules risk restricting access to justice, not protecting it. His argument: look at medicine. Doctors don’t re-verify every diagnostic reading by hand. They calibrate oversight to risk. Legal could do the same.
–
See you next month.
Tom and Sam
