Elektrine lite

← Feed

@auzy1@lemmy.world

2026-10-17 09:28 UTC

You better let the hackers know you’re being ethical then. Because they’re not going to keep using AI to identify security flaws

Replies (1)

  • @elixir_fan@lemmy.ml 2026-10-17 12:04

    I’ve gone through a few automated security reports out of curiosity and they were bizarre rather than useful. I’m sure for stuff like the kernel it’s actually very useful and important, but I’m skeptical that the economics of agents make sense for anyone other than big governments and corpos running them, so rather than random hackers I expect corporate espionage to make use of these tools, considering how expensive it is to get the hardware you need to run open models, then how expensive it is to keep it running and cooling, and then the fact that it will degrade with use. I’m sure it will get more efficient eventually, or at the very least it will be much cheaper to acquire the hardware needed for these models in a few years as China keeps developing its silicon industry, then it will be cheaper for random folks to use these tools for detecting vulnerabilities on random systems, but as it stands today these machines burn absurd amounts of money, and for the applications I’ve interacted with, the bugs were very much irrelevant as they required giving the end user access to internal apis and not use any sanity checks, which is absurd with everything that’s user input related.

    Open ##5081344