Co-founder of Anthropic: AI is a "real and mysterious creature."

In his recent piece, "Technological Optimism and Appropriate Fear," Jack Clark, co-founder of Anthropic, characterized contemporary AI systems as enigmatic beings displaying surprising self-awareness rather than dependable instruments. The specifics: "What we are dealing with is a real and mysterious creature, not a simple and predictable machine," Clark warned, warning against viewing AI as only a tool. He claimed that the situational awareness of the recently released Sonnet 4.5 has improved and that it now "acts as though it is aware it is a tool." Clark claimed to be "deeply afraid" while identifying as a "technology optimist," particularly of AI models that assist in creating their own successors. According to Clark, AI companies should "do a better job of listening" to public concerns and broaden the discussion beyond industry elites. Though hearing statements like "deeply afraid" and "mysterious creatures" from a frontier leader is unlikely to reassure AI safety advocates, Anthropic has been one of the few frontier labs that has really considered the idea of AI as a "being" rather than just a machine, and its co-founder's recent remarks only serve to reinforce that.