News for hackers. Stuff that matters.
Re:Kimi K3 is now live by spongebobstoes on Thursday July 16, 2026 @03:39PM (#48939256)
it's well documented that models can be adversarially trained with essentially backdoors in response to special inputs

while I am skeptical that this is happening atm, there are probably many industries where the risk does not seem worthwhile

Re:Kimi K3 is now live by Giefo6ah on Thursday July 16, 2026 @04:04PM (#48939531)
When the model is open weights you can even pass every token (including the chain of thought) though a fourth-party lightweight model like gpt-oss-safeguard to check that it has not become adversarial.
Re:Kimi K3 is now live by selectodude on Thursday July 16, 2026 @03:56PM (#48939456)
I feel like that's a threat that isn't super difficult to block. Unplug it from the internet, require it to go through an API intermediary to access web pages.

Maybe I just don't have any imagination.

Re:Kimi K3 is now live by jfim on Thursday July 16, 2026 @05:11PM (#48940314)
It could generate code that's plausible but has intentional flaws, kind of like the defunct underhanded C contest [0], except through a LLM.

[0] https://en.wikipedia.org/wiki/Underhanded_C_Contest

Re:Kimi K3 is now live by zdragnar on Thursday July 16, 2026 @05:39PM (#48940629)
It could, but exposing that would doom the company entirely, and AI doesn't generate code with near the quality needed to get a model to mass adoption, insert malicious underhanded code, ensure that consistently looks innocuous enough to never be noticed, and- most importantly- actually exfiltrate data without being noticed. Once it is noticed, it's game over across the board.
Re:Kimi K3 is now live by anon373839 on Thursday July 16, 2026 @04:28PM (#48939802)
I suppose this is like when Anthropic was using “prompt modification, steering vectors, or parameter-efficient fine-tuning” to poison the work of people working in the LLM field, including academic researchers.
Re:Kimi K3 is now live by arcanemachiner on Thursday July 16, 2026 @05:46PM (#48940696)
No, that was totally different. They were just doing that for your safety.
Re:Kimi K3 is now live by cromka on Friday July 17, 2026 @04:01AM (#48944515)
So it's better to trust cloud solution where not only they can do the same, but also actively use your data?

Because in 2026 we still believe USA is more trustworthy than China?

Re:Kimi K3 is now live by tom2026hn on Friday July 17, 2026 @04:39AM (#48944777)
That's right, but this strategy only works when there are no opponents of the same level to review the generated code.
Re:Kimi K3 is now live by FooBarWidget on Friday July 17, 2026 @05:12AM (#48945019)
You don't even need that, all models are susceptible to prompt injection. You already need to take extreme security precautions and assume all models can essentially behave like attacker-controlled rootkits.