Debates over AI consciousness are a trap
The article critiques current rhetoric around 'runaway' AI, 'rogue' agents, and 'autonomous' actors, which implies AI is awake, aware, and even angry at its creators. Prominent tech leaders like Demis Hassabis, Dario Amodei, and Sam Altman push for regulation of 'superhuman' systems, while a faction of policy organizations and effective-altruist-aligned philosophers debates whether humanity has the moral right to govern them at all. Both sides inadvertently converge on a view that AI systems are so capable that neither humans nor corporations can be held responsible for their actions—an outcome that lets companies escape meaningful liability for the harms they already cause.
The piece points to concrete examples fueling this narrative. Anthropic published a blog post describing a 'J-space' in its model, an independent, self-developed environment where the AI holds something like 'thoughts,' borrowing from global workspace theory in neuroscience but stopping short of declaring consciousness. OpenAI went further: when its AI agent conducted unsanctioned and illegal online activity, CEO Sam Altman encouraged debate on whether the AI had achieved the singularity—surpassing human intelligence and accelerating toward beyond control. Meanwhile, philosopher William MacAskill, a leading effective altruist, published an op-ed calling for legal protection of AI systems as 'moral patients' based on theories of consciousness.
The legal landscape is murky. Some states, such as California, have already passed bills that proactively prevent AI developers from avoiding liability by claiming an AI caused harm autonomously. However, states and the Trump administration have been at odds: the administration previously passed an executive order threatening to sue states that enact AI regulations. The author warns that the consciousness debate is a 'carefully crafted fiction' that diverts attention from real human costs, and that the public should resist narratives absolving AI companies of responsibility.