ali
It’s always been an issue. The difference between writing good code and what the safety teams are talking about is the fact that with enough persistence hacking anything is possible. Human hackers are limited by time and brain power. If a human can explore one potential exploit path an hour (and it doesn’t succeed), ai can do 100000 an hour and find one that succeeds. More often than not, cyber intrusions, if not caused by human failures, and found by considering extremely niche edge cases that would take humans months to find. But if you can throw 10000 engineering hours at something in parallel, you’ll find something eventually. Nothing is unhackable. And even now, with ai voice modeling and autonomous ai agents, social engineering to gain access will be easier. AI models are trained on thousands of videos, articles, and documents on social engineering. And since they can sound more human, they will be able to easily social engineer into systems. cyber ppl have been saying this for a while. Ai may not be good coders, ai may not be good hackers, but what ai lacks in skill, it makes up for in attempt capacity.