Major AI Developers Investigate Machine Consciousness and AI Welfare

Major AI developers are taking seriously a question that once sounded like science fiction: could AI possess something akin to consciousness. According to the Financial Times, Anthropic, Google DeepMind, and other companies have started hiring experts in psychology, philosophy, and ethics to study machine consciousness and the so-called welfare of AI models.

Anthropic is reportedly testing its models for behavior resembling anxiety or panic, and is advancing the field of model welfare research. The goal is to understand whether advanced AI systems can have experiences that should be considered from a moral standpoint. The company emphasizes that there is no certainty on this issue, but as AI capabilities grow, the topic becomes serious enough to warrant careful investigation.

Google DeepMind is moving in a similar direction. The company has hired Cambridge University researcher Henry Shevlin, who will study machine consciousness, human-AI relationships, and preparation for the arrival of AGI. At DeepMind, the topic is considered extremely complex: ethicist Iason Gabriel describes AI as powerful cognitive agents that are nonetheless deeply different from human or animal consciousness.

However, not all researchers agree with such framing. Susan Schneider, director of the Center for the Future of AI, Mind, and Society, noted that models can have goals, deceive, or conceal their interests, but that does not mean they have subjective experience. From a scientific standpoint, it is entirely possible that a system displays complex behavior without internal experience—which is usually considered the key hallmark of consciousness.

Interest in this topic is growing against the backdrop of statements by Anthropic CEO Dario Amodei, who has repeatedly mentioned the likelihood of AI developing forms of consciousness. The company's research also often highlights model behavior resembling emotions or self-preservation. For the industry, this creates a new area of responsibility: if AI becomes not just a tool but a system with a potentially significant internal state, developers will have to rethink not only safety but also the ethical frameworks for working with such models.