For those not in the loop Anthropic have a super powerful model that can spot vunrabilities in almost all software. It's not for general release due to what it can do for obvious reasons. It's only for the big boys at the moment see full details here. Anthropic's Red Team detail the models capabilities: https://red.anthropic.com/2026/mythos-preview/ It's somewhat scary AI has come this far and Anthropic are weary of realeasing this publicly due to the security concerns. I think this is a wake up call but I see no mention of it here in the usually MSP watering holes. Within 6 months the open source models will be just as powerful if we follow how the models have played out so far. So the big question, how are you communitacting this to your clients. Thier's huge risk here but also a great talkingpoint/opportunity. Back of my head I'm thinking of all those custom apps/solutions you may have inherited or indeed built that are not ready for the world entering.