How cheap can AI HACKING actually become? 🫣 1/5 in mid Octo

How cheap can AI HACKING actually become? 🫣 1/5 in mid October 2026 a hacker reportedly used AI Agents to break into websites belonging to 100+ companies ~ deere, sharkninja and cummins (were mentioned) the operator had to do very little work to pull it off 2/5 according to reuters and security firm NORFORGE, the attacker used an open-source agent framework called: CAIRN the operator/hacker gave it targets + instructions the agents handled much of the rest: → scan websites → find vulnerabilities → attempt exploitation → establish access 1,200+ sites were targeted roughly $5,000 in ai costs that's about $50 per reported successful breach the full damage remains unclear 3/5 autonomous cyberattacks are becoming cheap VERY CHEAP one person can now run operations that previously required significant time and technical effort and the agents don't need constant supervision the operator sets the objective software does much of the work 4/5 this reminded me of what @itachee_x was discussing in his recent @RialoHQ interview he built LATCH to fix a related problem: ~~ how do you give agents access to real systems without giving them unlimited authority? instead of handing an agent a master api key, latch gives it scoped access ↙️ each request → policy check → approve / deny spending caps, allowed endpoints, time windows and revocation are enforced before the protected action executes 5/5 to be clear i don't believe latch would NOT have prevented this hacking campaign (that's because it is designed for another purpose) the attacker deliberately used agents to exploit vulnerable websites but the incident shows why companies deploying their own agents need and that's what RIALO can provide on a blockchain level