
First Apple M5 memory exploit discovered using Anthropic AI, gives root access on MacOS
https://www.tomshardware.com/tech-industry/cyber-security/apple-m5-architecture-suffers-first-privilege-escalation-exploit-anthropics-claude-mythos-helps-researchers-bypass-memory-integrity-enforcement

12 Comments
All this is in the end good for companies to fix their shit. AI will lead to super secure hardware and software
I’m so glad we’re sacrificing the environment and power grid for this.
[The CIA agent that’s been losing all his tools but still had a 0 day for the new macs](https://youtube.com/watch?v=8wVFvtowR6I)
The interesting part isn’t that M5 has a vulnerability — every chip does. It’s that AI found it faster than humans would have. Security research is becoming an arms race between AI-assisted attackers and AI-assisted defenders, and whoever has the better model wins.
This will lead to better software overall, and that’s better for everyone.
that’s pretty wild if it holds up 😭 AI being used to find exploits that quickly is kinda scary and impressive at the same time
Security like everything costs, time, convenience , not exploiting shortcuts … just no one wants to pay the freight . Now it’s surprised picachu
Does the exploit require physical access to the machine?
Double edged sword.
Everyone is thinking that AI found it but doesn’t read the first sentence.
The gap in security was found by AI _assisted_ security researchers
Are you eligible for the bounty money if you’ve used AI to discover an exploit?
Until the full technical details are made public and / or reproduced by independent researchers, I call BS. This is marketing hype until then.