Saw this interesting blog post today about how the NHS is apparently preparing to shut down all Open Source repos in response to models like Mythos posing a security risk. Sad to see this response from the NHS.
Interestingly, the Linux kernel vulnerability I posted the other day was partially identified by AI. I can see why there’s a kneejerk reaction to close repos, but think it demonstrates a misunderstanding of cyber-security to move ahead with the idea… The code’s already out there; the models have already trained on it. I doubt at this stage that the NHS is going to vastly restructure their enormous codebases from the ground up. You don’t even need source code access necessarily to exploit vulnerabilities – and with this, only NHS devs can contribute to the defence and security of their code.
Seems like a very costly move for very little benefit.
What are your thoughts? Do you think we’ll see more of this sort of thing as frontier models improve?