It is fixable.
Model requires proof that you are a legitimate developer of that piece of software.
Every Anthropic/OpenAI account will have a list of projects the model is allowed to work on for security issues.
Model requires proof that you are a legitimate developer of that piece of software.
Every Anthropic/OpenAI account will have a list of projects the model is allowed to work on for security issues.
> A subsequent investigation found that the campaign to insert the backdoor into the XZ Utils project was a culmination of over two years of effort, starting in 2021, by a user going by the name "Jia Tan". They used sock puppetry in a pressure campaign against the original maintainer of XZ Utils, eventually being given maintainer permissions on the project.
As with clever, careful serial killers, it's tough to count the ones we haven't caught.
It's possible there are infiltrators who are still working on long-term infiltration and haven't yet attempted to add any malicious code anywhere, but the point is that in terms of actual attempts, we've seen a single one and it wasn't even successful despite years of prep.
No, we can't, as that happens a lot via non-serial killers.
A truly successful serial killer is likely one who hides in that noise. No taunting the cops, distributed geographic locations, random methods, avoiding calling cards, and careful not to leave too many traces.
It seems likely that some of the 350k unsolved homicides in the US can be explained this way.
> It's possible there are infiltrators who are still working on long-term infiltration and haven't yet attempted to add any malicious code anywhere…
Or the code's already there, latent, as it would've been in the XZ case, which got discovered by chance and someone very dedicated to looking into a performance glitch.
Since we do not know the ratio to undiscovered this "1-2" is meaningless to assess the risk of this sort of attack.
If the acceptance criteria is “would prevent every single past instance and every imaginable future instance”, then yes, no mitigation is every sufficient to address any problem in the world, so we might as well give up.
But I don’t think that’s the right lens to use.
You're in control of how much danger of accident you expose yourself to.
Nobody is in control of how much danger we are exposed to from other people who are actively trying to do us harm, who will keep going until they get what they're after or are stopped.
For most people, seatbelts are the former. Yeah, not perfect, but they reduce risk. For the latter, if you're known to be a seatbelt wearer, the attacker just does something where seatbelts don't matter.
Every new AI model introduces new capabilities and competencies, so we're not even sure what the true risk levels are yet for self-exposure in this category. The restrictions on AI may be like seatbelts and speed limits, or they may be like "if you install a 1000 HP turbojet engine in your Honda Civic it will no longer be road legal". And this analogy also includes how the first cars had speed limits set low enough to not risk the horse industry, i.e. we may be too cautious.
But I think people misinterpret the principle to mean that only perfect solutions have any value.
When in reality defense in depth is the opposite principle: you scan incoming emails for phishing, and that’s good but imperfect. You do mandatory training, and that’s good but imperfect. You use RBAC to limit blast radius, and that’s good but imperfect. And so on.
Among tech people, especially on HN for some reason, there’s this odd thinking style that goes: 1) company announces security measure, 2) the measure could possibly limit my freedom to do whatever I want with the company’s products, 3) I don’t like that, 4) I can come up with scenarios where the security measure is not sufficient, entirely on its own, to address the claimed risk, 5) therefore the security measure does no good at all, 6) therefore this is a PR smokescreen to disguise their desire to capriciously fuck with me out of pure malice, and I am angry about it
The Linux Kernel is in its training data. I just tested it. I copied about 20 random lines from the linux kernel and asked which codebase this was from and it could immediately tell.
Being able to attribute the source of a line of code doesn't help you to know if a repository can be legitimately hacked on.
As you could imagine, I might just take all or part of the Linux USB stack from the kernel to retrofit it into my own kernel.
For package X, I should be able to present my npm (homebrew, apt, nuget, etc) credentials with publishing rights for the package.
If package X is of sufficient public interest (user count, nature/sensitivity of user data, downstream distribution, etc), then the public interest + cryptographic credentials should permit access to best-available security auditing.
Yes, we still are trusting trust, that the owner of the package itself is not malicious, but that's not a sharp degradation from status quo.
If you try to do some kind of dupe-detection, someone can use a lightweight LLM to make superficial changes until it's considered a different project.
Finally, the meatspace status quo is that it is totally acceptable to pay someone to find security bugs in someone else's open-source software, such as the Linux kernel.
Even if you don't, a lot of source code can be legitimately copied thanks to the GPL/MIT/BSD/etc. I'm allowed to take all of zlib and integrate it into my own project if I so chose.
Your private fork doesn't meet the conditions described.
Presumably your ID so that feds may pay you a visit when they feel like it, your email need not apply.
I’m surprised that there’s even enough pushback against ID verification to matter, all the corpos are probably salivating at the idea of having fully accurate profiles of everyone, think of the ad and product targeting. The govt. would also love that, for different reasons.
It’s not too hard to imagine a future where you can only use certain things only with the govt. mandated spyware installed - bank apps already often don’t work on rooted Android phones (and you’re expected to use those apps to confirm payments) and all sorts of certification exam software is basically that already if you take a test remotely.
It follows that the same principle would just get pushed further, like what Discord wanted to do etc. Same with how Apple requires your documents for a developer account, Hetzner for a hosting account or Twitch for getting paid by them and tax stuff.