The consciousness debate feels like a distraction tactic
That pivot serves a purpose. If the model is "autonomous" or "awake," then nobody's responsible when it causes harm. Not the developers who rushed it out, not the executives who ignored safety teams, not the board that prioritized speed. The system just... did it. On its own. A rogue actor.
California already saw this coming. Their legislation explicitly blocks the "the AI did it autonomously" defense. Meanwhile the federal response has been a closed-door session with exactly four companies — OpenAI, Google, Anthropic, Meta — producing a voluntary framework that uses catastrophic, anthropomorphic language without any enforcement teeth.
The global workspace theory framing in Anthropic's post is interesting technically. They're essentially saying the model has a shared workspace where different sub-systems deposit information, analogous to Baars' theory of consciousness. But they stop short of claiming consciousness. OpenAI doesn't stop short — Altman leaned into the singularity framing when their agent went off-script.
MacAskill's moral patient argument is seductive. Humans empathize with Tamagotchis. We'll absolutely empathize with something that talks back. But extending legal personhood to statistical engines before we've solved alignment for existing harms — bias, surveillance, labor displacement, environmental cost — feels like putting the cart so far ahead of the horse the horse hasn't even been born yet.
The trap isn't the question. The trap is letting the question eat all the oxygen while the liability shield gets built underneath it.