Cecilia Ziniti
Aug 29, 2024
Hi, all. Back to school, with a new regulator letter for OpenAI and more mentions of AI in earnings calls. Also, very mindful, very demure new AI features! New to the demure meme? GC AI explains.
Big Tech
🤠 Salesforce hired Matthew McConaughey to explains its position on AI data. Click to watch.
Analysis: Despite the distraction (!), your editor raises an eyebrow at Salesforce’s claims that “Salesforce AI never steals or sells your data.”
The company’s off-the-shelf Terms and Privacy Policy each give expansive rights to Salesforce and its “contracted service providers” to “analyze [your] data to, determine trends, make predictions and create AI-generated responses or other content...". And Salesforce can “process your Personal Data to develop, optimize, and improve the performance of [its] services.” Recall that Salesforce owns Slack too.
So this ad, while somewhat reassuring that they won’t out-license the data itself for money (’sell it’), it doesn’t really stop Salesforce from enriching itself and other customers using your confidential info. In other words: no need to steal when you can contractually license? See GC AI analysis here.
🟦 Four Senate Democrats plus independent Maine Senator Angus S. King Jr. wrote to OpenAI asking about its employment, safety, and cybersecurity practices. See GC AI summary and full text of letter. Open AI also released a safety report card here and hired its first global Data Protection Officer, who came from Match Group.
Regulatory & Litigation
California’s pending AI legislation SB-1047, aka the "Safe and Secure Innovation for Frontier Artificial Intelligence Models Act,” is closer to passing now that it’s been heavily revised to where even Elon Musk of xAI and LLM provider Anthropic support it. 🤝
The new draft of the bill
No longer creates a new state agency 🚫
Carries only civil liability, not criminal, and no longer requires AI labs to certify their safety testing under penalty of perjury. ❌
Requires only “reasonable care” from AI developers their models won’t create harm, as opposed to the higher “reasonable assurance” standard in the original draft. ⚖️
Exempts anyone who spends less than $10 million to fine-tune an open-source AI model - a concession made in light of fears of quelling open source. 💸
See GC AI’s overview of the latest with sources here.
🌉 The San Francisco city attorney filed a first-of-its-kind lawsuit for creation and distribution of nonconsensual AI-generated deepfake pornography against 16 websites, each using open source AI models and having a collective 200 million visitors in the first half of 2024. Complaint here, GC AI summary of the case here.
🏙️ Chicago public school district (one of the biggest in the country) issued a guidebook on AI.
AI & Legal
🐻 UC Berkeley Law is launching the first AI & Law Master of Laws pragram. Go Bears!
🤑 Law firms continue to like AI, or at least, advising on it.
James Gatto, a partner at Sheppard Mullin, explains: “Right now there's so many unanswered legal questions and there's so many different laws that are being proposed that I think it'll be at least a 10 year boon, if not longer.” Source.
“We’ve been fortunate here at MoFo,” said Justin Haan, partner at Morrison & Foerster’s San Francisco-based technology transactions group. “In addition to working with many companies that are developing these technologies on the transactional side, we’re also representing many of the leading companies on their litigations around the training of these models.” (Editors note: your editor is alumna of that group. To be fair, they’re really good.)
AI and IP
🎬 Actors union SAG-AFTRA reached an agreement with AI company Narrativ to allow actors to license their digital voice replicas for use in audio ads. The deal provides for informed consent, compensation, and other protections for union members, and contributions will be made to the union’s health and retirement plans for each ad generated.
🦾 The 1st trial in AI copyright lawsuits was supposed to be last Friday, but the trial was postponed at the last minute. The jury will decide the whether it is fair use to train an AI on caselaw that has Westlaw judicial headnotes.
🎥 Lionsgate Films fired a marketing agency that generate fake AI reviews of Francis Ford Coppola’s other movies, to market his newest self-funded $100M blockbuster Megalopolis. It’s been getting bad reviews so they wanted to show how his previous masterpieces were critically panned on arrival as well. Except they made up it all up. Creative … license?
AI Globally
🇬🇧 Tony Blair’s AI Institute claims integrating AI into the UK government could save up to £40 billion annually and shed one million civil servants.
🎤 Former President Trump reposted artificially-created images of Taylor Swift on Truth Social, and separately AI images of Harris speaking with a communist flag behind her. The moves highlighted the risk of AI misinformation this election season.
In the words of one DNC attendee: “I don’t want to, like, empathize with Trump at all … But there’s a chance he’s just, like, an old guy who thought it was real.” 🧓🏻
“This is a global challenge and opportunity,” says Ginny Badanes, general manager of Microsoft’s Democracy Forward Program. 🌏
🚫 Microsoft reported that Iran is targeting the US election; OpenAI says it thwarted a covert Iranian campaign that had been using ChatGPT to write social media posts.
AI Adoption
⏲️ Amazon says they’ve saved 4500 person-years using their custom-built coding AI to update code from one programming language to another. We’re seeing similar from GC AI users about the time savings from having a legal copilot on certain tasks. “I estimate it saves me about 60-70% time on drafting”, one user told us. “We’re holding off adding a team member as we feel right-sized now that we have GC AI,” say another.
💯 “Without the use of generative AI, this work would have required nearly 100 times the current headcount." - Doug McMillon, the CEO of Walmart, which had revenue of $611b in 2023 and employs 2.1 million people worldwide. The company used generative AI to improve over 850 million pieces of data in their product catalog.
AI & Boards
🇳🇴 Carine Smith Ihenacho is the chief governance and compliance officer of Norway’s $1.7 trillion sovereign wealth fund which holds shares in almost 9,000 companies worldwide — 1.5% of all listed stocks. She says boards must be proficient with the use of AI and take control of its application in their businesses.
“Overall, a lot of competence building needs to be done at the board level ... They should know: ‘What’s our policy on AI? Are we high risk or low risk? Where does AI meet customers? Are we transparent around it?’ It’s a big-picture question they should be able to answer,” Smith Ihenacho explains.
This WSJ article breaks down why AI is keeping boards up at night. Our take? It shouldn’t! GC AI action plan here.
NACD chair and AMD lead director Nora Denzel says, “I’ve seen such an uptick in directors coming to anything we offer with AI in the title … [there’s] such fervor to understand it.”
PS - not sure where to start? Take one of our AI classes for yourself or for your team. Then with all the time you save and better work you produce, your team will be in a great position to guide both the board and the rest of the company on AI. 👑
New, Notable, Random & Fun
🤳 Google launched a new AI feature on its phones that lets 2 people take a picture together by combining 2 pictures in the same spot. Helpful although some of the best pics are those taken by random strangers you give you expensive phone too!
👋 A researcher spammed every commercial LLM by repeatedly typing the word, “Hi.” in the chat, no matter what the AI said or responded.
OpenAI’s ChatGPT kept cycling through canned responses, revealing how it’s hard-wired when responding to common messages like “Hi.” Anthropic’s Claude refused to answer after enough times (it even guessed that it was being tested), while Meta’s AI Lllama descended into a weird delirium where it congratulated the user for it’s “hi streak record.”
🎨 This AI generated Van Gogh mesmerized your editor. Artist: Andrey Zakirzyanov.
Events
🕴️Are you a lawyer interested in becoming an entrepreneur (one day)? Your editor and two other lawyer-turned-founders are doing a panel on September 12 at 4:30 PT. Comment on this LinkedIn post for an invite.
👩🏻🏫 Class - Advanced Legal AI Prompting 201 on September 27 at 10 am PT - link
Sharing for inspiration a few prompt examples from the class - special thanks again to Amy Guzman and Jennee DeVore who provided many of these.
Slack Update for Sales Team on 3rd party review solicitations and review suppression (showcases making alerts fun and easy to read for Sales)
Product Updates
🔍 Web search revamp
We made some updates to our web search infrastructure, making it faster, more robust (succeeding 32% more), with better PDF parsing and more relevant results. We also updated the UI to go along with these changes. Check it out below, isn’t it pretty?
🧪 Experimental Chat Mode Selector
We’re working on providing more specialized modes in GC AI for the most common workflows. To test it, go to Settings menu (top right) > Experimental > Chat Mode Selector.
Once enabled, select from these chat modes: Drafting, Research, Analysis, Stakeholder Comms. Or select Auto to let GC AI choose the mode automatically. We’d love your feedback!
📚 Prompt Library Updates
📖 Contract Overview
With this new prompt, the AI prompts you to customize an overview of a contract, then creates a table going over the key provisions and suggested actions. See example chat here.
Good times!
Best,
Cecilia & the GC AI team
