He removed a ton of images that were requested to be removed.

A lot of models you don't know what's in their dataset, so they might be 'weird.' His dataset is opensource, so it can be understood.

"We believe that Open is required for auditable models, for private data, IP-rich data, and there's a business model that's amazing for that." "Especially if you're talking about regulated industry."

He said you can have both. Closed models for things that's good for, and Open for private data.

In all their models they have invisible watermarking, and they have attribution coming soon.

"50% of code on GitHub is AI-generated now."

He signed the FLI paper with Musk and others.