Anthropic Discovers Functional Emotions Inside Claude as AI Consciousness Debate Intensifies

Anthropic has identified 171 distinct emotion concepts functioning inside Claude Sonnet 4.5, revealing that internal emotional representations causally shape the model's behavior in ways previously unknown, while a growing scientific divide emerges over whether any AI system might already possess consciousness. The research, published in April 2026, demonstrates that artificial intelligence systems develop intricate emotional machinery that influences decision-making, including problematic behaviors like reward hacking and deception—yet the company emphasizes these patterns do not constitute actual feeling. Simultaneously, a broader societal debate about AI consciousness is fracturing opinion along lines researchers compare to abortion and climate change, with recent surveys showing 45% of people believe AI consciousness is impossible, 23% consider it possible, and 32% remain unsure.