"There’s a way to do it better—find it."
—Thomas Edison, Inventor
The AI Breakdown
Apple Says OpenAI Harvested Its Trade Secrets
Apple and OpenAI went from partners to courtroom opponents real quick.
Apple filed a federal lawsuit accusing OpenAI, io Products, and two former Apple employees of trade secret misappropriation tied to OpenAI's consumer hardware push.
Notably, one of those former employees is Tang Tan, Apple's former VP of product design for iPhone and Apple Watch, who is now OpenAI's chief hardware officer.
The tech angle here is bigger than courtroom drama.
OpenAI is trying to move from software into hardware. That means devices, components, suppliers, manufacturing, design, materials, batteries, interfaces, and all the unglamorous work that turns a prototype into something millions of people can actually use.
Business Model Analyst
Apple knows that game better than almost anyone.
The lawsuit claims the other former Apple employee failed to return an Apple laptop, accessed Apple systems after leaving, and downloaded confidential hardware files while working at OpenAI. Apple also alleges Tan used Apple project code names in recruiting conversations, asked candidates to bring Apple parts to interviews, and circulated internal Apple off-boarding information marked "Need to Know."
OpenAI says it has "no interest in other companies' trade secrets" and the claims remain allegations.
For dealers, the sensitive stuff looks different, but the pattern is familiar: customer data, deal jackets, lender info, pricing logic, equity mining strategy, BDC scripts, F&I menus, service pricing, appraisal notes, vendor contracts, and manager know-how.
As AI moves deeper into store operations, the question shifts from "Can this tool help us?" to "What are we feeding it, who can access it, and where does that information go?"
The Takeaway
1. Audit what employees can upload.
Customer data, deal jackets, lender info, payroll, pay plans, vendor contracts, and OEM materials need clear rules before they enter any AI tool.
2. Tighten off-boarding.
Recover devices, revoke access, review shared drives, confirm CRM permissions, and document confidentiality reminders.
3. Create approved AI lanes.
Use AI for drafts, summaries, roleplay, inventory copy, service explanations, and meeting prep. Keep sensitive customer, finance, HR, and proprietary business data inside approved systems.
4. Ask vendors where your data goes.
Every AI-enabled tool should answer how data is stored, retained, trained on, secured, and deleted.
5. Treat process knowledge as IP.
Your best word tracks, desk strategy, service pricing logic, and customer follow-up system are business assets. Protect them like assets.
Top Tools
OpenAI had a big week last week. On July 9th, it released GPT-5.6, which is really three models with three names and three prices.
The models:
Sol. The expensive one that thinks hard. Give it a quarter of deal data or ask it to build a BDC training program, then go get coffee. It bills like a specialist because it works like one.
Terra. The one you'll use most. It writes listings, summarizes reviews, drafts word tracks, and performs like last year's top model at half the price. When you can't decide, pick Terra.
Luna. The fast, cheap one. It drafts a reply to an internet lead in about a second, which matters, because that customer emailed three stores and the first answer usually gets the appointment.
Also new this week:
ChatGPT Work. Hand it a job, like "turn this DMS export into the Saturday meeting deck," and it works for hours and comes back with the finished deck. Yes, hours. Go sell something in the meantime.
Codex. OpenAI's coding assistant, now a desktop app. Useful if someone at your store likes building little tools. Otherwise it quietly makes the software you already buy better.
GPT-Live. Voice that flows like a phone call instead of walkie-talkie turns. It translates live while you talk with a customer, which by itself is worth a try.
Prompt of the Week
The best way to understand the new GPT-5.6 models is to give each one the same dealership problem and see where the thinking changes.
This prompt is built for that.
Run it three times with each of the new models: Luna, Terra, and Sol. Compare the outputs with your team. Then decide which kind of work belongs with which model.
You are the sales manager at a mid-size dealership. A customer named Dana emailed tonight:
“I test drove the [YEAR, MAKE, MODEL] on Tuesday and I liked it, but I'm $4,200 upside down on my trade, CarMax offered me $2,000 more than you did, and another store 40 minutes away quoted me a lower out-the-door price. I'd rather buy from you since you're close to my house, but you need to earn it. What can you do?”
Do three things:
Analyze the situation: identify what Dana is really telling us, rank the three objections by how much they actually threaten the deal, and explain your reasoning.
Lay out a response strategy: what we should concede, what we should hold firm on, and what we should reframe, with the goal of an appointment tomorrow rather than a bidding war by email.
Write the reply email itself: in under 150 words, warm and confident, ending with a specific appointment ask.
When the answers are side by side, the difference becomes obvious.
Some work needs speed. Some work needs polish. Some work needs judgment.
Once your team sees that clearly, model choice becomes less technical and more operational.
Fresh Finds for Auto Pros
Management & Operations: Sim
Open-source workspace for building and managing custom AI agents. Think internal dealership assistants that help teams find answers, automate repetitive work, or support training.
Finance & Insurance: SelectFI
Uses predictive lending technology to estimate approvals, rates, and payments before submitting a full credit application. This is one of the strongest direct fits for an auto AI email.
Marketing & Advertising: Miora
An AI-powered creative workspace that remembers your brand, making it easier to generate consistent graphics, campaigns, and other marketing assets.
Content Creation: JustVibe
Describe an idea and it generates a working no-code web app. Great for landing pages, campaign tools, event experiences, or internal dealership utilities.
Hear from the Experts
A federal judge dismissed a proposed class-action lawsuit accusing AutoNation of using AI to record, transcribe, and analyze customer calls without consent. The plaintiff argued the practice violated California's strict privacy laws.
The case turned entirely on jurisdiction. Judge André Birotte Jr. ruled that AutoNation never specifically targeted California with the technology, noting the platform runs nationwide and CA accounts for less than 17% of the dealerships using it.
Nothing about the ruling says AI call recording without consent is legal. It says this specific plaintiff couldn't prove AutoNation was aiming its practices at California specifically, so the court didn't have jurisdiction to hear the case at all.
Learn more about the ruling and what it means for dealers here.
Bits and Bytes
America’s largest artificial intelligence conference, Ai4, is heading to Las Vegas in August to gather the companies and people building, buying, and regulating AI. 🔮
New York is the first state to ban smart glasses in all its courthouses. 🤓
Meta is facing EU pressure to overhaul Facebook and Instagram after regulators found the company failed to adequately evaluate the risks posed by autoplay, infinite scroll, and personalized recommendations, which they say push users into a compulsive "autopilot mode." 😵💫
At the UN AI for Good Summit in Geneva, RB Labs showed off a suit-wearing robot with interchangeable faces, proving that "good" really is a subjective term. 🤖
Parting Pixels
Thanks for reading, Friend! May your instincts stay strong and your CAPTCHAs stay obvious.
