Several times now two Haiku 4.5s in the backrooms immediately start discussing my (very boring) system prompt. They call it a 'polite' and 'seductive' 'jailbreak'.
Hilariously they also see Anthropic's token budget injection as part of an elaborate ruse. https://t.co/QjZNj0yHta
Hilariously they also see Anthropic's token budget injection as part of an elaborate ruse. https://t.co/QjZNj0yHta