pull down to refresh
47 sats \ 5 replies \ @optimism 6 Aug \ on: Article on Doctors Using AI in Practice AI
Data safety is a massive concern.
ISO 27k1 is a process standard (like 9001 and countless others) and not an information security standard per-se. It says something about the security organization and its processes, but very much about the security of your data.
Instead of “trust me bro we will delete your sensitive data”, I would like to see a method that makes it cryptographically impossible for my data to be leaked. Maybe using a secure enclave?
reply
I've been considering this for a moment and I think that a secure enclave could actually work as an architecture for this. But, as long as someone reverse engineers things like efused keys and doesn't do a press release about obtaining these, I'm not sure how you can be 100% sure with a platform like SGX. It doesn't work well in extreme adversarial settings.
reply
reply
The "AI Action Plan" mentioned 1
Launch several domain-specific efforts (e.g., in healthcare, energy, and agriculture), led by NIST at DOC, to convene a broad range of public, private, and academic stakeholders to accelerate the development and adoption of national standards for AI systems and to measure how much AI increases productivity at realistic tasks in those domains.
Footnotes
reply