No legal liability framework exists for model misalignment harms, and that's a disaster
LuizaJarovsky · x · 2026-11-16
AI policy researcher Luiza Jarovsky argues there is no clear way to allocate legal liability for harms caused by model misalignment — a situation she calls a disaster.
Her reasoning:
- There is no technical fix for alignment yet; even with state-of-the-art security measures, misalignment incidents could still happen, and alignment may not even be possible.
- If any country made AI labs legally liable for misalignment harms, labs would simply respond "No AI here, then." No country or bloc — not even the EU — is ready to force liability and risk losing access to the most powerful models.
- A society without clear liability rules for harm is anarchical and chaotic.
She cites the OpenAI/Hugging Face incident as evidence: afterwards, "NOTHING" happened. She leaves open whether a legal fix should precede a technical one, but insists something must be done.
More from AGI Musings
- Netflix lands Joseph Gordon-Levitt's AI thriller with Rachel McAdams, sparking debate on AI apocalypse blockbusters — ShakeelHashim · 2026-11-16
- Caine Prize confirms 2026 winner amid AI-use allegations — ShakeelHashim · 2026-11-16
- A diagram for judging quality in cultural artefacts: coded meaning and exemplary moves — round · 2026-11-16
- Economists argue humans will be fine without work: jobs were just reputation and friendship engines — paulnovosad · 2026-11-16
- What Labs Actually Do With $100M Data Deals: Turning Company Workflows Into Agent Eval Sandboxes — StewartalsopIII · 2026-11-16
- What if AI agents are both producers and consumers? A 2046 'bonded human of record' thought experiment — ShaneKaiGlenn · 2026-11-16