GenAI Ads Debut in the NBA finals!

Hi all, welcome to this week’s update!

The future of ads just made it to the NBA Finals.
In only 48 hours, Kalshi used AI to create a commercial that aired on ABC, cutting costs by 95 percent. The result? A bit unhinged, but it shows how fast AI can go from prompt to primetime for only $2,000.
I thought big brands would hesitate before trusting GenAI at this scale, but this proves some are ready to move fast, even when the output feels a bit chaotic.

In today’s AI for the 9 to 5:

• 📈 Deleted Chats Aren’t Gone: AI Privacy Takes a Hit

• 🎤 AI Ads Hit Primetime: Kalshi’s NBA Finals Experiment

• 🎤 Meta Just Bought a Piece of the AI Supply Chain

• 📚 WWDC: Is Apple Taking an AI Gap Year

🚀 Deleted Chats Aren’t Gone: AI Privacy Takes a Hit

OpenAI is now being forced to retain every user conversation, even deleted ones, after a court order tied to its legal battle with The New York Times. This applies to all ChatGPT users.

The New York Times argued that users might be infringing on its content and then deleting chats to hide evidence. In response, Sam Altman called the request "inappropriate" and warned it sets a bad precedent, even floating the idea of "AI privilege" similar to doctor-patient confidentiality.

Why this matters: As millions share sensitive information with AI, knowing that even deleted chats might be stored threatens user trust. And as AI starts moving into always-on physical devices, the privacy stakes keep rising.

🎤  AI Ads Hit Primetime: Kalshi’s NBA Finals Experiment

Prediction market Kalshi just ran one of the first AI-generated ads during Game 3 of the NBA Finals on ABC. The 30-second “unhinged” spot was built using clips created with Google’s Veo 3 video model.

The process:

• AI filmmaker PJ Accetturo made the full ad in just 2 days, generating around 300 to 400 Veo 3 clips to assemble 15 final shots.

• He shared his full workflow on X, using Gemini and ChatGPT for ideation, scriptwriting, and prompt crafting.

• The ad used Veo 3’s speaking capabilities, though there were still some issues with random subtitles and inconsistent voices.

• Accetturo said the whole thing cost around 95% less than a traditional production and predicted that “high-dopamine Veo 3 videos will be the ad trend of 2025.”

Why this matters:
This feels like a turning point for AI in advertising. The style is still a bit chaotic, and it won’t fit every brand, but the speed and cost savings are hard to ignore. And with video models improving fast, even the current rough edges may not be around much longer. I can imagine GenAI Ads will have a huge potential to improve the results from SMB ads in the next months.

🏢 Meta Just Bought a Piece of the AI Supply Chain

Meta is not sitting around waiting for GPUs to fall from the sky. Last week, they acquired a stake in Scale AI, one of the leaders in data labeling and AI model training infrastructure.

The details:

• Scale AI’s new funding round values the company at $13.8B, with Meta coming in as one of the investors.

• Scale helps companies like OpenAI, Anthropic, and now Meta, clean, label, and prepare massive datasets to train models.

• The deal gives Meta more access to high-quality training data, which is becoming one of the biggest bottlenecks in model development.

• Data is the new oil, and Meta just secured a refinery.

Why this matters:
Everyone is chasing GPUs, but high-quality data is the other side of the AI race. As models get bigger, clean data becomes more critical than compute. Meta’s move signals that owning the data pipeline may be just as valuable as owning the chips.

📚WWDC: Is Apple Taking an AI Gap Year

Apple kicked off WWDC 2025, rolling out a bunch of design tweaks, new names, and customization updates across its products. But when it comes to AI, the announcements felt pretty thin, especially in a year where everyone else is going all-in.

Here’s what dropped:

• Live Translation is now built into Messages, FaceTime, and calls, processing everything locally on-device to protect privacy.

• Visual Intelligence lets you analyze whatever’s on your screen, search for similar products, or even ask ChatGPT questions about images.

• Shortcuts now gets AI-powered automations, with ChatGPT integrations to help generate workflows.

• Developers get access to Apple’s on-device AI models through a new framework, avoiding cloud API costs.

• Workout Buddy launches on Apple Watch, using AI to give real-time personalized coaching based on your biometric data.

Why this matters:
Sure, Apple added some AI features, but they feel more like side dishes than the main course. In a year where every other tech giant is dropping full AI product lines, Apple seems to be intentionally keeping it light. The talk of this being an AI gap year for Apple feels pretty accurate, and it’s a weird vibe for a company usually known for setting the pace.