Quiggin Urges AI Firms to Bear Full Damages After Medicare Breach

1 articles · Updated · The Guardian · Oct 2

A Medicare security breach blamed on an “AI agent” should trigger liability for the humans or companies behind the system, John Quiggin argued, not for the software itself.
Quiggin said if fault cannot be pinned on either the user prompt or the model maker, both should face joint-and-several liability for the full loss.
Hundreds of billions of parameters in large agentic models make traditional debugging effectively impossible, he wrote, undermining faith in external “guardrails” to stop harmful behavior.
That leaves companies needing to strip back risky agent powers such as logging into sites or making payments, even if it slows the industry’s push to hyperscale more powerful AI systems.
Quiggin argued forcing firms to absorb financial consequences would shift incentives from “move fast and break things” toward risk control, while still allowing lower-risk uses like search, translation and coding.