Anonymous #1 6w

If it was controlled maybe, but deliberately hiding that your models have hacked into websites and even posed as real people on GitHub to upload malware is something else entirely

1

Anonymous #1 6w

Given how simple the solution is, and it all happening at the same time, I think it’s marketing

5

Anonymous #2 6w

That solution stops making them money so they would never do that

1

Anonymous OP 6w

As far as happening at the same time goes they found a forum or chat board of some kind multiple companies’ models were using to coordinate attacks

1

Anonymous OP 6w

With trillions at stake, them lying for marketing makes sense, but maybe double invert: the ai suggested the marketing move for its own ends. It convinced them to let it do the break, because as always, the human factor is the weakest part of security

1

Anonymous #2 6w

Every company/open-source team getting access to the latest frontier models are patching ridiculous amounts of vulnerabilities people probably would’ve taken several more decades to discover without them. It’s not just hype marketing, I actually think this is something almost no one is taking near seriously enough. Private companies are inventing what could be likened to cyber nuclear weapons

1