Post #2522099

@kentpitman@climatejustice.social @vnikolov@ieji.de @bagder@mastodon.social When you said that the LLM vulnerability scanning turning up similar vulnerabilities and bugs to previously discovered bugs, that company that was in the news for the 2025 vulnerability discoveries in curl and openssl somewhere did have a page saying they specifically asked their AI to look for code that was similar to commits introducing previously discovered vulnerabilities. So there is an aspect to LLMs looking for their keys under the streetlamp.