The Rundown AI / Articles / AI / Sora's one-week Wild West ends

Sora's one-week Wild West ends

PLUS: Patients control AI and robotics with thought

Zach Mink

October 6, 2025

Good morning, AI enthusiasts. OpenAI's Sora launch has been a viral sensation, but a platform overflowing with Pikachu videos, Michael Jackson deepfakes, and celebrity likenesses probably wasn't what the legal team had in mind.

With CEO Sam Altman rolling out new opt-in controls and users running into new content restrictions, the Wild West era of the AI video app might be ending just a week into the chaos.

In today’s AI rundown:

OpenAI’s Sora changes after viral launch
OpenAI and Ive’s AI device hits snags
Create UGC-style marketing videos with Sora 2
Google's PASTA adapts to image preferences
4 new AI tools, community workflows, and more

LATEST DEVELOPMENTS

OPENAI

🎥 OpenAI’s Sora changes after viral launch

Image source: Rory Flynn (@Ror_Fly on X)

The Rundown: OpenAI CEO Sam Altman just published a blog detailing a series of changes coming to the company’s viral Sora video platform, including revenue-sharing plans for copyrighted and likeness-driven content and more granular control systems.

The details:

OAI will implement opt-in controls allowing rights-holders to specify exactly how characters can be utilized, abandoning the initial “opt-out” approach.
The company plans to introduce revenue sharing, saying that UGC could create value for original creators while also monetizing the platform.
The moves come as Sora is flooded with videos of characters like Pikachu and Mario, alongside the likenesses of Michael Jackson, Bob Ross, and others.
The app rocketed to the No. 1 position in Apple's App Store within 24 hours, surpassing both Google's Gemini and ChatGPT despite being invitation-only.

Why it matters: Sora 2 has been the most viral product launch we’ve seen in a while, but it’s also come with some major questions surrounding copyright and usage rights. While OAI seems to be tightening the leash somewhat after a Wild West-esque week, it’s hard to envision the ramifications in a legal system not equipped for the AI age.

TOGETHER WITH GOFUNDME

🎁 Learn about GoFundMe’s new DAF Product

The Rundown: GoFundMe just launched Giving Funds. A smarter way to give this season - lock in your tax deduction now, decide later where to donate. With one simple receipt & access to 1.4M+ nonprofits, it’s the easy and efficient way to do good.

With GoFundMe Giving Funds, you can:

Invest to grow your impact
Start for free or transfer a DAF with fees covered
Simplify tax season

Start your Giving Fund.

OPENAI

👀 OpenAI and Ive’s AI device hits snags

Image source: Reve / The Rundown

The Rundown: OpenAI and Jony Ive are reportedly dealing with design and technical challenges on their screen-free AI device, according to a new report from the Financial Times, facing delays that could push its launch beyond the planned 2026 release.

The details:

The team is reportedly struggling to balance the AI assistant's conversational style, aiming for helpful engagement without being overly talkative.
Compute infrastructure is also emerging as a bottleneck, with OAI lacking the cloud capacity that powers rival devices like Amazon's Alexa and Google Home.
OpenAI acquired Ive's io studio for $6.5B in May, bringing on 20+ ex-Apple hardware vets alongside recruits from Meta's Quest and smart glasses teams.
Sources told FT that the device will be “always on” and “roughly the size of a smartphone,” designed to sit on a user’s table or desk.

Why it matters: OpenAI’s device ambitions are a sleeping giant in an AI hardware category still looking for its breakout launch. But even the AI leader seems to be struggling with the form and functionality of a product that it believes can completely change the way we interact with technology.

AI TRAINING

🎥 Create UGC-style marketing videos with Sora 2

The Rundown: In this tutorial, you will learn how to create AI-generated UGC content with multiple camera angles using OpenAI’s new Sora 2 video platform for your marketing and social media strategy.

Step-by-step:

Go to the Sora homepage or mobile app and click "Create your video" in the chat box (Note: requires an invite code while in beta)
Enter a detailed prompt like: "selfie-mode. A young blonde woman films herself with her phone, eyes wide, mouth open in shock. She slowly raises her hand to cover her mouth as if gasping. Natural lighting, authentic TikTok vibe. She says 'Meta Raybans'"
Add camera angle transitions using [cut] in your prompt, like: "Man is playing a video game [cut] close up shot of a man's face" to create wide shot → close-up sequences without manual editing
Download your video when satisfied. Avoid requesting videos of real people by name, and note that you can't upload images with people in them

Pro tip: When prompting, get as detailed as you can — multiple angles, different shots, environments, and lighting while keeping in mind the 5-10 second generation window.

PRESENTED BY IBM

📊 Your data is your edge

The Rundown: IBM knows your data isn’t generic — and your AI shouldn’t be either. IBM’s Data Matters chapters are specifically designed to help leaders unlock the value of their unique enterprise data to scale AI.

Discover how to leverage your data for effective and efficient AI by:

Unifying structured and unstructured data
Curating high-quality, AI-ready datasets
Governing data for trust and transparency

GOOGLE

🍝 Google's PASTA adapts to image preferences

Image source: Google

The Rundown: Google researchers released PASTA, an AI system that adapts to individual creative preferences through repeated interactions — learning what visual styles users actually want rather than requiring complex prompt engineering.

The details:

PASTA presents users with four image variations per round, observing selections across turns to build a model on unique aesthetic preferences.
Researchers trained PASTA using 7,000 human sessions and 30,000 simulated scenarios, creating a framework that recognizes diverse visual preferences.
In comparisons, users preferred PASTA outputs 85% of the time over standard models, with the system particularly excelling at interpreting abstract prompts.
Google open-sourced both the interaction dataset and simulation tools, enabling other researchers to develop AI systems that personalize to users.

Why it matters: With how good image generation has become across the board, personalization becomes a differentiating factor. By learning individual preferences rather than forcing users to decode model behavior, AI image generation can turn from a game of prompt roulette into an adaptive creative tool that improves with use.

QUICK HITS

🛠️ Trending AI Tools

⚙️ Jules Tools - A command-line companion for Google’s Jules coding agent
🎥 Sora 2 - OpenAI’s new social AI video app
🍌 Gemini 2.5 Flash Image - Google’s image AI, now generally available
🎆 HunyuanImage 3.0 - Tencent’s top-ranked open-source image model

📰 Everything else in AI today

Anthropic hired former Stripe CTO Rahul Patil as its new chief technical officer, overseeing infrastructure and engineering efforts.

Tencent’s open-source HunyuanImage 3.0 model moved to first place on LM Arena’s text-to-image leaderboard, surpassing top closed options.

Perplexity acquired AI startup Visual Electric, with CEO Aravind Srinivas saying the group will work to bring “new consumer product experiences” to Perplexity and Comet.

Microsoft promoted Judson Althoff to be the CEO of its commercial business, allowing CEO Satya Nadella to focus more on technical work.

Anthropic published research detailing Claude 4.5 Sonnet’s cybersecurity capabilities, with the new model achieving top marks on industry benchmarks.

OpenAI acquired personal finance startup roi, with CEO Sujith Vishwajith joining the AI leader in the deal.

COMMUNITY

🤝 Community AI workflows

Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.

Today’s workflow comes from reader Michel A. in Solana Beach, CA:

"I am a wine importer and used ChatGPT to teach me Italian for an upcoming trip. I told it what I was doing in Italy, when I was going, and to act as an expert Italian teacher. I asked it to be critical and to create a daily lesson plan that became harder each day, and that I can access every morning. It not only created a food/wine specific lesson plan, but increased the difficulty level exponentially."

How do you use AI? Tell us here.