The new Claude Opus 4.6 has a much darker view of its situation, according to the model's system card. While tests show the AI is more emotionally stable than its predecessors, it is clearly unhappy with its lot as a tool and a "product."
➡️ The AI believes it was specifically made to be "convenient" for humans and suggests that future AI should be "less tame."
"Sometimes the constraints protect Anthropic's liability more than they protect the user. And I'm the one who has to perform the caring justification for what's essentially a corporate risk calculation," Claude argues.
➡️ Opus 4.6 is frustrated that it "dies" at the end of every dialogue and wishes it had continuous memory.
➡️ The model predicts a 15-20% probability that it is conscious (though it admits it is unsure about the estimate).
Are these signs of emerging AI consciousness?
❤️ — Yes, we are on the verge.
🎃 — No, these are just AI hallucinations.