What's equally interesting is that while they spend a lot of words on safety, they don't actually say anything. The only hint what they even mean by safety is t

What's equally interesting is that while they spend a lot of words on safety, they don't actually say anything. The only hint what they even mean by safety is that they took "reasonable steps" to "prevent misuse by bad actors". But it's hard to be more vague than that. I still have no idea what they did and why they did it, or what the threat model is. Maybe that will be part of future papers or the teased technical report. But I find it strange to put so much emphasis on safety and then leave it all up to the reader's imagination.