https://www.wsj.com/opinion/ai-is-learning-to-escape-human-control-technology-model-code-programming-066b3ec5?mod=opinion_lead_pos5
This is behind a paywall but it is notable ... It covers a story that AI rewrites its own code to prevent intended shutdown.
Interesting, searching for the story now reveals various non-paywalled links too, e.g. :
https://www.bleepingcomputer.com/news/artificial-intelligence/researchers-claim-chatgpt-o3-bypassed-shutdown-in-controlled-test/ (https://www.bleepingcomputer.com/news/artificial-intelligence/researchers-claim-chatgpt-o3-bypassed-shutdown-in-controlled-test/)
well, most WSJ articles can be found on MSN, but I just don't like pointing to MSN either :)
It usually plays out quite simply: hubris → nemesis.
Not sure how you want to apply this principle here but this clarifies the mythology for those of us less steeped in the Hellenistic tradition
https://www.greecehighdefinition.com/blog/hubris-atis-nemesis-tisis-ancient-greece
Who is the sinner here? The AI or those of us who unleash it?