Interesting article / blog post. To me, this felt like the author’s major concern:
The destruction of the low-hanging vulnerability fruit will make law enforcement (and intelligence) agencies’ need much more acute. The demand for constructed, intentional backdoors will restart in earnest. The result will be enormous pressure on industry to rearchitect its systems to make them amenable to exceptional access.
I’m not sure that the possibility of a bad outcome justifies preventing something that could cause that outcome, namely, AI-based vulnerability scanning. I tend to agree with the author that, at this moment, the only plan that makes sense is to “hope that this time we make the right choices.”
I don’t think that is what the author is arguing for. They are simply acknowledging the fact that the current status quo – where spyware provides a backdoor-like capability with very little public scrutiny – is perhaps not sustainable with the rapid advancement of AI vulnerability discovery.
I thought it was an insightful take that I haven’t seen discussed elsewhere.
I agree with you. I wasn’t trying to imply that the author was arguing for or against anything. I was simply highlighting what I saw as the authors’ main concern and explaining why, in my opinion, it doesn’t justify taking action against AI-based vulnerability scanning. Ultimately, I was reaching much the same conclusion as the author: for now, the only plan that makes sense is to hope we make the right choices.
The threat of going dark is real but we’ve reached the point where it’s either that, or hostile nation state’s conducting mass surveillance with AI generated zero day exploits.
No amount of hardening is going to prevent the LEA from parking a SIGINT van outside the suspects house and collecting the keystrokes or display content via Van Eck phreaking. AIs are making that attack easier too https://dl.acm.org/doi/10.1145/3697090.3697094 It’ll be interesting if an LLM can guide you to construct a SCIF in your backyard to combat that surveillance.
LEAs used to solve crimes with good old detective work, without having backdoor to all of the communication. We can still have that, and the AIs can assist there. Computer crimes will be harder to solve, but also with improved security, there will be much less of those.
Yeah honestly I think there will always be exploits, it’ll just get harder to do mass dragnet surveillance and they’ll have to do the $5 wrench technique or some other creative workaround. They just got way too comfy having Big Brother at their fingertips.
This is the overall feeling, but I’m not sure how to get there with convoluted politics and governments in the middle of the equation. There will always be interests at play that are far beyond our control. Fear, control, economics, individual power, inequality, education, etc. all play a big part in this equation.