242 | 😘 😂 Thanks wife.

In today’s Brainyacts:

1. Delaware court has best AI use policy so far

2. SPECIAL SEGMENT: Clifford Chance’s Copilot Use Cases & ROI

3. Siri powered by ChatGPT is coming next week!

4. Not much AI model news today

5. Did AI Chatbot lead to a suicide? and more news you can use

   👋 to all subscribers!

To read previous editions, click here.

Lead Memo

🏛️ 👍🏻 Delaware Court chooses wisely on AI governance

Delaware Supreme Court shows us a pragmatic, logical, and common-sense approach to guiding the use of GenAI as it adopts interim policy providing guidance on its use by judicial officers and court personnel.

Here is the policy language:

1. Authorized User Remains Responsible. Any use of GenAI output is ultimately the responsibility of the Authorized User. Authorized Users are responsible to ensure the accuracy of all work product and must use caution when relying on the output of GenAI.

2. Informed Use. Authorized Users should not use Approved GenAI without a working knowledge and understanding of the tools. Authorized Users should be trained in the technical capabilities and limitations of Approved GenAI prior to use. 

3. Decision Making. Authorized Users may not delegate their decision-making function to Approved GenAI.  

4. Compliance with Laws and Judicial Branch Policies. Use of GenAI must comply with all applicable laws and judicial branch policies. 

5. Non-Approved GenAI. Authorized Users may not input any Non-Public Information into Non-Approved GenAI. Non-Approved GenAI may not be used on State Technology Resources.  

Click here for the Court’s Statement and here for the Order.

This no-nonsense policy that holds users accountable for their choices, reminding us that no matter how polished or persuasive an AI’s output may seem, the responsibility for accuracy and reliability rests squarely on us. This policy brings a refreshing counter to the widespread tendency to overly govern the use of AI—highlighting that these models are not infallible sources of truth.

We have learned that most models’ confident tone and conversational ease lure users into thinking it’s perfectly accurate, but the Court’s policy cuts through this misconception. It simply doesn’t let users hand off their decision-making duties or blindly trust AI without a deep understanding of its limitations. Instead, it sets out clear boundaries: training, informed use, adherence to existing legal and ethical standards, and stringent data privacy. This framework is durable and sustainable, giving AI a role that supports rather than supplants human judgment.

Special Segment. I am trying to be fair.

Recently, in 241 | 🎛️ the “Precision Prompt”—a simple tool to fine-tune your interactions I talked smack about Microsoft’s Copilot (I do this reguarly). I still think its strategy is harmful to people struggling to see the value in using Generative AI. In this instance, I took issue with the fact that in its recent announcement, Microsoft used Clifford Chance as an example of success using Copilot but gave no details. Well, I found some details.

Thanks to Legal IT Insider, we get a bit more insight. Clifford Chance shared its journey with generative AI and how it finally selected Copilot to test and adopt. Here are the use cases and ROI they expanded on. Here I break it down for you.

Use Cases for Copilot

1. Summarization

• Description: Copilot can provide summaries of extensive email chains or document exchanges, streamlining client reporting. Although output requires verification, this feature simplifies digesting large amounts of information quickly.

• Benefit: Saves time on manual summarization, improves communication efficiency, and enhances client updates.

2. Information Retrieval (Finding the “Needle in the Haystack”)

• Description: Enables quick searching within SharePoint and email to locate specific information, such as details on previous transactions (e.g., bond issues).

• Benefit: Reduces time spent on information retrieval, improving response times and productivity.

3. Originating and Structuring Information

• Description: Assists with tracking and organizing responses or structuring witness data. For example, identifying which banks in a syndicate have responded to an inquiry.

• Benefit: Enhances organizational efficiency by automating the tracking and structuring of critical data, saving substantial administrative time.

4. Document Analysis and Creativity Support

• Description: Assists users in evaluating arguments and identifying weaknesses in documents, supporting more robust document creation and editing.

• Benefit: Supports higher-quality document preparation, enabling a more robust review process and strengthening argumentation.

5. Self-Service and Empowerment

• Description: Facilitates self-service options for staff, reducing dependency on service desks and increasing self-sufficiency regarding compliance and policy-related queries.

• Benefit: Reduces service desk workload and encourages employee autonomy, decreasing internal support costs and promoting knowledge empowerment.

Speculated ROI

According to the reporter, the ROI on Copilot primarily hinges on time savings and productivity gains:

• Cost Estimate: At $30 per user per month, for approximately 7,000 employees, Clifford Chance’s Copilot investment amounts to around $2.7 million per year.

• Break-even Point: Clifford Chance estimates that if each lawyer saves just 15 minutes per week, the Copilot investment would pay for itself. With expected time savings potentially exceeding 15 minutes, this suggests a positive ROI.

• Additional Gains: Beyond direct cost recovery, Copilot enhances document quality, accelerates data retrieval, and reduces dependency on support services. The expanded self-service options and streamlined reporting capabilities imply further savings and productivity improvements across multiple practice areas, ultimately maximizing value from the investment.

Thanks again to Clifford Chance and Legal IT Insider for sharing these details.

Spotlight

🗣️ 📲 Get ready for Siri powered by ChatGPT

Apple Intelligence is coming soon. The release of iOS 18.1 should happen next week. Now there will likely be a waitlist (yes, waitlist) to get access to Apple Intelligence but it should move fast.

For those users who choose to use the new Apple Intelligence capabilities, here is a breakdown on how to use Siri powered by ChatGPT and what happens to your data along the way. 

Getting Started with Apple Intelligence in Siri (iOS 18.2)

1. Control Your ChatGPT Access

After updating to iOS 18.2, ChatGPT integration is off by default, giving you the power to turn it on when you’re ready. You decide which features—like Writing Tools, Visual Intelligence, and Camera Control—Siri can access through ChatGPT.

2. Choose When Siri Uses ChatGPT

When enabled, Siri won’t automatically call ChatGPT unless you ask. For complex queries, Siri may suggest ChatGPT if it seems beneficial, but you’ll be prompted to confirm. You can turn off this “suggestion prompt” in settings, so ChatGPT is only used on direct request.

3. Privacy in Different Scenarios

Here’s how your data is handled, depending on your account setup:

• No Linked ChatGPT Account: Requests are anonymized and stripped of identifying data—Apple sends only the request and any attached files, like documents or photos, to OpenAI. Your IP is masked to further protect your privacy, and OpenAI is prevented from using your data to improve its models. Apple shares limited data with OpenAI, including only your location to help detect fraud, but all data remains detached from your Apple ID.

• With a Linked ChatGPT Account: When linked, you can access ChatGPT’s paid features directly through Siri. OpenAI’s data policies apply here, so some requests, attachments, and session details are logged to improve its models. However, none of this data is linked to your Apple ID, keeping your Apple account separate from OpenAI’s data. For both cases, Apple collects only general data—such as request counts and responses—to operate the service and ensure its security.

4. Integrated Visual and Writing Tools

Siri’s ChatGPT integration extends beyond Q&A. With Visual Intelligence, you can ask Siri to generate images, tapping directly into ChatGPT’s image generation features.

AI Model Notables

► OpenAI whistleblower disgusted that his job was to vacuum up copyrighted data to train its models.

► xAI, Elon Musk’ AI company is seeking to build AI Agents

► Google releases a method to detect AI-generated text at scale. Is watermarking text a path forward to tell what words have been written by a human versus machine?

► Apple launched a $1M bug bounty ahead of its major AI cloud release next week, offering rewards to security researchers who can successfully hack and find vulnerabilities in its private AI infrastructure.

► When your new junior associate is the legal AI agent - meet Alexi. https://www.alexi.com/

► Google's $2 Billion Anthropic investment faces heightened U.K. antitrust scrutiny.

News You Can Use:

➭ The White House issued a national security memorandum directing federal agencies to accelerate AI adoption. It emphasizes its use in sensitive government areas like defense and intelligence.

➭ Generative AI is making e-discovery practices more niche and specialized favoring boutiques over BigLaw.

➭ Mother sues tech company after she claims that her son’s “relationship” with AI chatbot drove son to suicide.

➭ In response to 👆, Common Sense Media published a parents' guide to AI companions and relationships.

➭ More than 11,000 actors and artists signed a statement labeling the unlicensed use of creative works for artificial intelligence training a “major unjust threat” to creators.

➭ Chipotle launched a new conversational AI hiring platform called ‘Ava Cado,’ which the restaurant says can accelerate the hiring process by up to 75%.

➭ Yahoo, McAfee team up to verify authenticity of news images.

Who is the author, Josh Kubicki?

Some of you know me. Others do not. Here is a short intro. I am a lawyer, entrepreneur, and teacher. I have transformed legal practices and built multi-million dollar businesses. Not a theorist, I am an applied researcher and former Chief Strategy Officer, recognized by Fast Company and Bloomberg Law for my unique work. Through this newsletter, I offer you pragmatic insights into leveraging AI to inform and improve your daily life in legal services.