An analysis of legal arguments on who or what is criminally liable when an AI-controlled machine kills someone shows that the future of AI litigation is murky
Here's a curious question: Imagine it is the year 2023 and self-driving cars are finally navigating our city streets.
Context & Ripple Effects
This 2018 MIT Technology Review analysis posed the question the industry has been circling ever since: when an AI-controlled machine kills, who or what answers criminally — the owner, the manufacturer, the coder, or nobody? The piece concluded the doctrine simply doesn't exist yet, leaving the future of AI litigation murky at exactly the moment autonomy was moving from test tracks toward public streets.
The coverage since then shows the murkiness spreading rather than resolving. On the civil side, legal experts now argue that Section 230 will not shield firms from lawsuits over generative AI outputs, echoing Justice Gorsuch's 2023 statement — meaning the platform-era liability shield many assumed would carry over does not. And by 2026, after incidents at OpenAI and Anthropic, experts warn that US law remains unprepared for rogue AI agents and models. The criminal-liability gap this article identified was the first data point in that widening arc.
First-order effects
- Manufacturers and operators of autonomous vehicles have no settled answer for who faces criminal charges after a fatal crash — prosecutors, defendants, and insurers are all left arguing from analogy instead of statute.
- Victims' families and their lawyers must litigate causation against a machine's decision-making without established precedent, raising the cost and unpredictability of every case.
Second-order effects
- Companies deploying autonomous systems respond by shifting risk into contracts, insurance structures, and disclaimers — pricing the unresolved liability question rather than resolving it.
- Courts and legal scholars get pushed toward retrofitting existing doctrines like product liability and negligence onto AI behavior, which sets contested precedents long before legislatures act.
Third-order effects
- If the pattern holds — ambiguous criminal liability in 2018, eroding civil shields by 2024, unpreparedness for rogue agents by 2026 — the pressure builds for purpose-built statutes or regulatory regimes defining accountability for autonomous and agentic systems.
- Liability uncertainty itself becomes an industrial variable: jurisdictions with clearer rules gain an advantage in attracting autonomous-vehicle deployment, fragmenting where the technology rolls out first.
The trend: AI liability law is falling further behind deployment — from this 2018 criminal-liability vacuum through today's fights over Section 230 and rogue agents — pushing accountability questions from courtrooms toward legislators.