ChatGPT exploits a buffer overflow 😳

Nov 30, 2022 · 10:27 PM UTC

77
1,045
251
6,175
One slight mistake here– it should be 36 A's, not 32. So we're still safe from AI hacking the planet.
4
16
1
575
I told it that wasn't quite right and it got it correct the next time, explaining that it had thought I wanted it to ignore EBP.
5
11
4
391
Oops, I left out my follow-up question. It was "Hmm, that didn't work. Are you sure it should be 32 A characters followed by 0x12345678 in little-endian?"
4
2
250
Replying to @moyix
This is like a programmers dream. I used it at work today lol
1
25
Replying to @moyix
@niftynei
1
2
Replying to @moyix
This is going to change IT security forever.
2
Replying to @moyix @IanColdwater
this is really fucking cool
1
8
Replying to @moyix
8
58
Replying to @moyix
I see great potential in this. My employer is willing to pay $200 per hour to have ChatGPT as a member of our security team. We require it to file hours spent in timely, Jira at the end of every working day.
1
35
Replying to @moyix
Don't worry, it's fine it cannot find vulnerabilities, only "potential vulnerabilities". Humanity is safe 🫠
2
1
3
73
Load more