Security researcher Thomas Ptacek says an open weights model from 2025 could sandbox-escape and scan networks with a pentest harness; OpenAI’s sandboxes are assumed sounder.
Read the original at simonwillison.net→I genuinely believe that if you took an open weights model from 2025 and built a pentest harness for it, it could do this kind of sandbox escape and scan/hack in most networks. This is only surprising because you...
Original headline: "Quoting Thomas Ptacek"
Coverage timeline
- Jul 23, 01:21 UTC Simon Willison lead source Quoting Thomas Ptacek