The AI & Legal Update #7 - 🍎 Apple Intelligence

Cecilia Ziniti

Jun 16, 2024

Hi, friends. Apple joins our AI party, Adobe stock surges even after people hate on their legal terms, and California looks to regulate AI. Long update today.

Read time: 9 minutes, 08 seconds.

Special Feature - Apple Intelligence Announcement

🍎 Apple developed a small, low-power AI model for use on-device within iOS’s native applications. It'll summarize, rewrite, and proofread. Apple calls it, “AI for the rest of us.” For more complex questions that Apple's own AI can't answer, Apple will rely on ChatGPT.

Privacy-wise, Siri will request permission to call ChatGPT from users on a per-question basis. Your editor’s product counsel take? Give people a global setting, like Steve Jobs recommends ("privacy means people know what they are signing up for ... ask and ask again... make them tell you to stop asking.").

Most exciting? iOS will get the full slate of unicode emoji responses for tapbacks in iMessage. That's a 700x upgrade (!) over the current 5 colorless tapbacks (🩶,👍,👎,“haha”, !!, and ?) which have left your editor feeling constrained.

🤝 Also note, OpenAI and Apple’s deal is a barter, no-money deal. Apple says billions of users are valuable reach, which, fair. Wall Street liked it, with Apple gaining $329B this week - more than the value of OpenAI ($80B), Anthropic ($18B) and, well, all other AI startups combined (h/t Axios).

Apple also markets better privacy for its AI

Apple went hard on marketing privacy for AI, even turning its logo into a lock (weird by ok). The company says it won’t share user data with OpenAI.

But! Apple still doesn’t give you access to your own Siri interaction history, like Alexa and Google Home have for a decade. Is Apple really better? Their designer of the new apple-to-lock animation above thinks so. Then again, if you have to say it …

Elon Musk disagrees Apple's doing privacy right

🤠 Elon Musk tweeted that “Apple has no clue what’s actually going on once they hand your data over to OpenAI”, and that "Apple devices will be banned at [his] companies [as] an unacceptable security violation.” OpenAI responded that they “care deeply about user privacy.” Elon was otherwise happy this week because shareholders approved his record-breaking pay package (see GC AI summary of the proxy vote and of the original Chancery Court ruling).

OpenAI & Other Big Tech

GPT-5 is coming + new OAI execs are here

🔎 Open AI has started training its next model, presumably GPT-5. We can’t wait! They appointed two new execs, CFO Sarah Friar (former CEO of OpenDoor) and Chief Product Officer Kevin Weil (ex-Twitter), and filled its board cybersecurity seat with retired former NSA director and ex-US Army General Paul M. Nakasone.

🏗️ Altman cements his position

Remember OpenAI ex-Board member Helen Toner’s big accusations? Sam Altman disagrees - click the image to watch the clip.

Commentators see Altman cementing control - getting himself back on the board, adding board members, and pushing them to adopt a more standard corporate structure for OpenAI. Speaking of, Elon Musk just withdrew his ridiculous lawsuit that OpenAI breached a non-existent contract with him in creating a for-profit subsidiary.

Adobe learns the people read pop-ups about AI.

📈 Adobe’s stock popped last week after they announced better-than-expected results on the strength of their AI offerings in Photoshop and other products.

But week before last? Different story. The internet was furious after Adobe products popped up below.

Product Counsel take: it's sloppy to pop up context-less legalese in an AI world. Here’s my and GC AI's fixed version. But of course Adobe needs to view content for compliance. Like most Big Tech, Adobe’s tools are used by everyone for everything, including, say, by terrorist organizations for murder videos. Users eventually defended Adobe - it has always been in their terms (!) - and Adobe blogged about it and eventually revised its terms in response.

PS - we agree with GC AI user and tech company DPO Carl Gottlieb that the best way to do terms of use is with a concurrent blog post or FAQs.

Consumer & Litigation

🫠 Remember Microsoft’s litigation magnet AI feature, Recall, that would snapshot your screen every few seconds? After outcry, they’re making it opt-in and delaying it. Better than … recalling it? 🥁

🐾 An AI company Clearview allegedly unlawfully collected biometric facial geometry without consent. The company would have gone bankrupt to pay a litigation settlement, so … it’s giving plaintiff class members’ 23% of its stock. Interesting result from repeat internet class action plaintiff Jay Edelson. GC AI summary here.

⚖️ “[W]ith artificial intelligence raising the specter of lawless and limitless protections under [Section 230], we should revisit our precedent …” - Ninth Circuit concurrence in Cerise v. Meta Platforms (finding 230 did not bar claims for Meta's alleged breach of its own terms of use).

AI Adoption

📈 McKinsey’s latest report on AI found 65 percent of organizations are regularly using gen AI, almost 2X from ten months ago. Is yours one of them? 😉 (link to demo GC AI)

💵 PwC finds legal professionals with AI skills could will see wage premiums up to 49%. Sounds right. Folks using GC AI are getting promoted and getting more clients.

💻 A Reuters and Oxford survey found:

🦻 47% of Americans haven’t heard of ChatGPT.
🧮 7% use a chat AI like GPT or GC AI every day and 20% once a week.
🐥 AI usage skews young, with 56% of 18–24s say they have used ChatGPT at least once, compared to 16% of those aged 55 and over.

IP

🗞️ The New York Times’s 2023 copyright suit against Microsoft and OpenAI is in discovery now. See GC AI’s summary of the parties’ latest dispute. Armchair ruling? Umm… No way the number of custodians at Microsoft gets limited to 10 people in a case as complex as this.

🎗️ The USPTO will pay Accenture $70M for an AI prior art search tool.

Regulatory, Policy, & Litigation

🚗 Remember the story of LexisNexis and GM selling data collected from people's cars without consent? The Texas AG is looking into it.

🔌 The California Senate passed SB 1047, a bill that would require AI large model developers do various things like certify to a new “Frontier Model Division” regulatory arm of the California Department of Technology, and implement kill switches.

😱 VC Andreessen Horowitz says tech is shocked and blindsided by the SB-1047 bill, and many would prefer regulation of specific applications of AI, versus AI at large.

Editor’s take: this bill is aggressive. Pegging AI power to current math (models of a certain size) will get stale fast. Frontier Model Division sounds a little bit like … thought police? Your editor saw Senator Weiner live in SF last week. He seemed earnest enough but mostly spoke through a staffer and didn’t explain why the bill was needed. Couple takes below.

🕵 The FTC will probe Microsoft, while the DOJ will look into Nvidia on antitrust. FTC's Consumer Protection Director said, "if companies are making claims about their use of AI, or the capabilities of their AI, those claims have to be truthful and substantiated, or we are prepared to bring action.” He assures: “those are not just empty words." Hm. At GC AI we welcome reviews of our data on GC AI’s higher legal AI accuracy.AI & Law

📜 LexisNexis (of databroker and legal database fame) is acquiring a document drafting technology company, Henchman. Worst name ever for a legal company?

🌞 Law firms are starting to train summer associates on generative AI. “[J]unior lawyers won’t be replaced by AI … but they will need to harness it to be successful,” says a K&L Gates partner. Yep. PS - your editor went back to SF greek restaurant Kokkari this week, the very spot of her first summer associate lunch in the mid-2000s. Still delicious!

🥷 OpenAI reports on its efforts to stop covert influence operations using its tech. “Threat actors work across the internet. So do we.”

Random & Notable

🎹 Woah - Suno AI, who enables making cheesy AI pop, is working on AI to make a song from any sound. Beautiful accordion!
🚦Google’s project Green Light, an AI to adjust traffic light timing, reduced stop-and-go traffic more than 30 percent. See them live in Seattle.
💁‍♀️Dove released a beauty-related AI prompt playbook with some stats. AI draws women as blonde 37% of the time.
😨 AI apocalypse? ChatGPT, Claude and Perplexity all went down at the same time. We were uninterrupted at GC AI.

Events & Podcasts

👸 June 18 Open/Free AI Prompting Event with In-House Connect - Tuesday 9 am PT / Noon ET (free, 90 minutes) - hear prompting tips from legal AI expert and WGCN members Amy Guzman and Jennee DeVore with In-House Connect and moderated by your editor. Sign up here. This is a tweaked version of our popular class. Try it!

👩🏻‍🏫 June 21 - 101 Class - AI prompting for In-House Legal 101 class. 2 hours. 4.9

⭐. Sign up here. Discount for GC AI friends - code FRIENDS.

🎧 Hear your editor’s thoughts on legal AI with AI influencer Ate-a-Pi - here.