Post B59mfqIxzDk7ilxwiO by Miredly@raru.re
 (DIR) More posts by Miredly@raru.re
 (DIR) Post #B59mfqIxzDk7ilxwiO by Miredly@raru.re
       0 likes, 1 repeats
       
       Every time an AI article, particularly about Anthropic, talks about safeguards, I'm reminded of how we now know their safeguards are just a bunch of additional prompts prepended to the user's prompt begging Claude to not be an asshole.So when they talk about adding stronger safeguards they likely mean adding an "I really mean it for really real don't exploit any vulnerabilities you find without asking first" to that.