- Flaqueman@sh.itjust.worksEnglish5 days
When a headline is a question, the answer is always NO. Case and point
- lmr0x61@lemmy.mlEnglish5 days
Betteridge’s law my beloved
(It isn’t statistically true in practice, though 😔)
- 5 days
Microslop wants a lot of things… but they prove that they can’t deliver in term of quality and data safety.
- grue@lemmy.worldEnglish5 days
Literally adjacent in my feed:
Generative AI agents will never be secure; it’s a flaw inherent to their nature.
- Griffus@lemmy.zipEnglish5 days
You probably won’t see this, but I think you’ve gotten a response or two in your backlog.
- grue@lemmy.worldEnglish5 days
I regretted not cropping that as soon as I posted it because I knew someone would comment on it, but I couldn’t figure out how to crop after-the-fact on my phone and re-upload. The screenshot utility can do it, but the image viewer can’t.
- Griffus@lemmy.zipEnglish5 days
Sorry for being the one you saw coming, but I am now very fascinated that you can follow up on new ones.
- grue@lemmy.worldEnglish5 days
It’s mainly that I just don’t bother marking things read, so that’s like two and a half years of replies.
5 daysCopilot is the worst of them all. I wouldn’t trust it to do a grocery list let alone anything medical.
Microsoft’s AI wants to be your medical middleman, but is a “Secure by Design” promise really enough for Copilot? Would you trust Microsoft with the “puzzle” of your medical records?
Short answer? No, and no.
- fruitycoder@sh.itjust.worksEnglish5 days
Microsoft’s push to make Copilot a kind of AI medical middleman—especially through the newly announced Copilot Health—raises a real tension: the company is loudly promoting a Secure by Design philosophy, but the sensitivity of health data means the bar is far higher than a general security promise. The short version is that Secure by Design is necessary, but nowhere near sufficient for something that sits between you, your clinicians, your medical records, and your wearables.
- Microslop copilot
- docus@lemmy.dbzer0.comEnglish4 days
Security by design is only one aspect of what would be required. Even if it keeps my data secure, if it is going to recommend putting pva glue on cuts and butter on burns, it’s a no from me. Altough i would be curious what it has to say about vaccinations…
- phutatorius@lemmy.zipEnglish4 days
Security audit by independent third parties, including access to the full source code, or GTFO.
- fierysparrow89@lemmy.worldEnglish4 days
Source code of what? Unfortunately, none of the above is anywhere near enough.
We need locally available ai models that can run off-line. Also: the ai context and history must be kept separately from the model itself.
If the ai model needs to communicate with the outside world, user needs 100% transparency and control what data the ai sends.
- phutatorius@lemmy.zipEnglish4 days
Source code of what?
The AI agent. Also, a way to see all of its training data.
I don’t trust Microsoft with my temp folder, what makes you think I’m going to trust it with my medical data? In case there’s any ambiguity left in that: no, I do not, and will never, trust Microsoft with this data, nor with any other personal, personal adjacent, identifiable, personal, or private data. Period. Hard stop.
- HubertManne@piefed.socialEnglish5 days
thing is in a place with no universal healthcare there is a lot to lose. may be bad for places where healthcare is a right but at least its not going to completely prevent you from getting treatment.
