This whack-a-mole approach, while definitely good, is likely already dead as far as a way to prevent these actions. Local LLMs that have no restrictions will co

This whack-a-mole approach, while definitely good, is likely already dead as far as a way to prevent these actions. Local LLMs that have no restrictions will continue to get better. If anything the best thing about this post is not the actions they've taken, but simply that they've shown us a snapshot of what the future will look like for state-affiliated actors' actions. The research aspect I think is a good thing - giving people full access to more information for how things are made and architected will likely be a net positive. The phishing aspect though is terrifying - it's going to be crazy seeing what the next decade looks like for phishing. I do wonder how long it takes before there's some sort of "verified as a human" type function in communications to try and combat this.