Nvidia, Apple vs. OpenAI, and AI's Growing Pains

Dr. Ryan Ries here. Today I have 5 news stories that I think we need to discuss, covering a gamut of different topics. But before we do that, a couple of quick announcements:

We’ve got quite a few upcoming events you might be interested in.

Next week I’m taking off from Matrix for the holiday. I’ll be back on the 16^th.

1. Next week in Chicago on 9/8 – our last stop on the AMD / Mission Cloud roadshow.
2. Virtually, on 9/17 – I’m leading a session on how to find, score, and prioritize your AI initiatives. You’ll even gain access to a free custom app to do so.
3. NYC on 9/24 – Network, connect, and learn about the latest in agentic AI at our AI Innovators Summit co-hosted with AWS and Cloudera.

Now, here are the stories we’re covering:

1. Nvidia’s reported $12.9B purchase of Hugging Face
2. A startup’s claim that it made 19 million gallons of rain over Alaska
3. Apple & OpenAI drama – it’s heating up
4. Instagram’s crackdown on undisclosed AI profiles
5. OpenClaw 2.0 and what “multiplayer” AI coding means for your team

Nvidia is taking over the full stack

The Information reported that Nvidia agreed to buy Hugging Face for $12.9 billion. Neither company confirmed it publicly as of this writing, and no signed contract has surfaced, so treat this as a strong report rather than a done deal.

Hugging Face is the place where a massive share of the world's open-source AI models and datasets live. Developers download from it, fine-tune on it, and benchmark against it daily. It's closer to GitHub for AI than a typical startup.

Nvidia tried to buy a piece of it before. Last year Nvidia offered $500 million for a stake at a $7 billion valuation. Hugging Face turned it down, reportedly because it didn't want a single investor with that much sway over its direction. Now the number on the table is nearly double that valuation for the whole company.

My reaction:

GIF from GIPHY

With this deal, Nvidia is paying to control the front door of open-source AI. Hugging Face has stayed neutral ground until now. Models built for AMD chips, Google TPUs, and Nvidia GPUs all sit on the same platform, treated the same way. Once Nvidia owns the shelves, the question I’ll be asking is if neutral hosting can survive under one chipmaker’s roof.

A startup made 19 million gallons of rain??

Cloud-seeding startup Rainmaker flew drones over Alaska’s Kenai Peninsula in August, releasing silver iodide into clouds over seven missions across 3 hours. The company says the seeding produced about 19 million gallons of new precipitation, most of it snow, and that radar analysis lets them attribute that water directly to the flight.

The founder, Augustus Doricko, said this is the first time a private company validated cloud-seeding results with methods precise enough to isolate a specific volume of precipitation.

I’ll throw up one caution flag here:

The 19-million-gallon figure comes from modeling the difference between seeded and unseeded outcomes, not from measuring water that fell where no rain would have fallen otherwise. Modeled estimates and measured results are different claims.

But that doesn’t mean this technology is fake. Eventually the goal is to point these methods at areas like the Colorado River basin, where drought has strained the water supply for years.

Apple & OpenAI drama continues

Apple filed new evidence in its lawsuit against OpenAI, claiming a former Apple employee named Chang Liu used a confidential Apple circuit schematic in his work at OpenAI. Apple says Liu also used a tool sharing a name with an internal Apple engineering application, and that OpenAI knew about his access to Apple data.

Apple also claims Liu enlisted an OpenAI colleague to help destroy evidence in June, once he learned Apple was investigating him.

HBO, if you’re reading, can we get a drama series about this?

OpenAI’s defense so far centers on “residual access,” the idea that Liu only used old files after leaving Apple to help former coworkers who asked for his help, and that lingering system access after an employee departs is a common problem companies like Apple fail to manage properly. But Apple counters that Liu exploited a previously unknown authentication bug to keep his access alive… which is obviously different from a company forgetting to revoke his access.

Anyways, there is more to this story, and I found it to be an interesting read, so check it out if you’re into the drama.

Drawing the line with AI creators

Instagram announced its renaming its “AI creator” label to “AI-generated profile” and will reduce the reach of accounts that feature AI-generated people without disclosing it. Creators who apply the new label won’t be penalized for having an AI-generated person as their profile subject, but creators who skip the label will be penalized by the algorithm.

I like that Instagram is setting some boundaries here. The social platform found that users kept discovering profiles that looked human, and engaged with them like they were human, to only later find out the person never actually existed. A Wired investigation earlier this year found more than 2 dozen AI-generated male influencers promoting a dating app through direct messages to real users. And the New York Times separately found hundreds of AI-generated doctors and wellness personalities pushing health claims and supplements.

As someone who creates AI systems, if your company is running any customer-facing AI persona, chatbot avatar, virtual spokesperson, etc., this is a friendly reminder that disclosure standards are tightening across every platform, not just Instagram. Build your disclosure plan now rather than reacting to a policy change later.

OpenClaw 2.0: Turning a solo coding agent into a shared workspace

Remember the explosion of OpenClaw at the beginning of this year? Well, they released version 2.0 over the weekend.

The creator said this is the biggest update to the platform yet. OpenClaw is moving from a personal tool one developer runs locally toward shared infrastructure a whole team can use together.

So, what’s changed?

1. Shared sessions replace solo threads: Multiple people can now enter the same live agent session, see the same history, and add context without exporting or reconstructing what the agent already knows. One developer described handing off a project by simply joining a colleague’s existing session instead of writing a handoff document. The session became the handoff document.
2. A browser interface: Conversations, files, approvals, and live agent activity now sit in a shared workspace instead of scattered across a terminal. For our non-technical friends, this lowers the barrier to entry.
3. Security controls are more specific: Approvals can now tie to a particular command, session, and person. Sandboxing options expanded to match the isolation model used by rival tools like NanoClaw. None of this is on by default. OpenClaw’s baseline setup still assumes a trusted single operator and allows host-level execution unless an admin turns on stronger restrictions.

My Thoughts

AI is changing daily, but one thing that keeps coming up when I share these news stories is that AI’s capabilities keep outrunning the guardrails humans have built around it.

If you’re building or buying AI systems right now, here are a few questions I’d make sure to ask:

• Who controls the infrastructure layer your team depends on? What happens if that owner’s incentives shift?
• Does your organization have a plan for disclosure if you ever use an AI persona, chatbot avatar, or synthetic voice with customers?
• For any agent platform your team adopts, have you actually turned on the security settings, or are you running the default configuration and living on a prayer?

As always, build with intention. And if you need a partner with experience building these systems on AWS, we’re here to help. Reach out to our team here.

Until next time,
Ryan

Now, time for this week’s AI-generated image and the prompt I used to create it.

Create an image of me in a crab suit with a whole team of crab workers creating AI solutions on AWS. Use my reference photo attached.