AI Liability Is the Question Nobody Can Stop Asking — and Nobody Wants to Answer

A Florida campus tragedy is circulating this week with an uncomfortable question attached to it: can AI be held legally responsible? The thread isn't really about the technology. It's about the gap between the fluency of AI systems and the utter absence of anyone willing to own what they produce. That gap has become the defining fault in the AI ethics conversation right now — not "is this ethical?" but "when something goes wrong, who exactly is on the hook?"

The liability question keeps arriving through specific people in specific situations. A Pennsylvania judge sanctioned an attorney $5,000 for filing AI-hallucinated citations — for the second time — and the community reaction wasn't outrage at the AI. It was a kind of exhausted recognition that the word "ethics" is doing enormous labor in these moments, covering for a system where the humans closest to the output keep gesturing at the tool. One commenter put it plainly: the AI didn't file the brief. The lawyer did. The AI has no bar license to revoke.

Indigenous voices are making a version of this argument more pointedly.[¹] Where tech's mainstream ethics conversation tends toward frameworks and principles — "responsible AI governance," micro-credentialing, the language of certification — critics from communities with longer experience of institutional neglect are naming what's actually missing: no accountability, no checks and balances, no one who can be found when the harm arrives. The UN webinar circuit and the academic publishing world keep producing the vocabulary of ethical AI; the people who've watched institutions disappear when blame heads their way are skeptical that more vocabulary closes that distance.

What makes this moment distinct is where the conversation about responsibility is *not* happening. Regulatory frameworks are being cited everywhere — the EU AI Act, pending liability webinars, CPD courses — but the posts generating actual engagement aren't about policy architecture. They're about individual moments of consequence: an intern at a climate conference watching someone project AI-generated images while arguing that ethical AI is impossible, a student noting that "responsible AI" in university means adding "ir" to the first word before submitting, a commenter arguing that the money going into data centers should simply go toward improving human lives. These aren't policy proposals. They're expressions of a community that has absorbed the ethics vocabulary and found it insufficient.

The "Responsible AI" framework has spread so far that it no longer points anywhere. It appears in Pentagon summits, hospital systems, agricultural development projects, and charity academy curricula simultaneously — which means it has become a genre of institutional speech rather than a commitment. The most honest post in the current cycle came from someone who simply wrote that AI has little traceability to find what went wrong. Ghost in the machine. That's not a policy critique. It's a description of what accountability actually feels like from the outside — which is to say, it feels like nothing at all.