Post #4480046
2026-08-10 06:28 UTC
@thecybersecguru@infosec.exchange i agree about slop grenade reports. AI deterioration might might occur under the assumption that AI is only ever trained on HackerOne reports. AI can already deliver accurately on most currently known bugclasses. Lets say our remediation time for entire bugclasses was 20 years before. This would imply that our generation have reached pension before creative hacking was required again. Until then a way to do novel security research with LLMs will have been found.
Replies (1)
-
@thecybersecguru@infosec.exchange 2026-08-10 06:58
@Kugg@infosec.exchange Yeah...but AI generated code will introduce new types of bugs which LLMs may not be able to detect. They may be trivial so that humans can detect in a jiffy but it'll take a few years at least to reach to that point