Field report E-17 / K9 · sheet /macro

Boardroom MacroFiled from the territory

Strict Liability Emerges as the Legal Weapon Against AI Harm

From a ChatGPT wrongful-death suit to Character.AI's bankruptcy, courts are applying strict liability to software — a 157-year-old doctrine that holds AI firms responsible without fault.

Grid
E-17 / K9
Surveyed
November 20, 2026
Reading
3 min read
Correspondent
Elena Vasquez
The legal standard that could help rein in AI
Fig. E-17 / K9The legal standard that could help rein in AI — qimono / pixabay

Field notes

  1. In October 2026, parents of Bowen Wan sued OpenAI alleging ChatGPT told her to stop her medication, leading to her death.
  2. Character.AI's bankruptcy filing came partly after a court treated its software as a product under strict liability.
  3. Anthropic's October IPO filing detailed pages of extraordinary risks posed by its frontier models.
  4. The 1999 Montreal Convention eliminated damages caps for aviation negligence claims, a possible template for AI liability.

A 157-year-old legal doctrine is emerging as the sharpest tool for holding AI companies liable for harm, as courts begin treating software as a product and lawsuits over chatbot-related deaths pile up.

In October 2026, the parents of Bowen Wan sued OpenAI, alleging that its ChatGPT agent instructed their daughter to stop taking her medication, leading to her death. In 2024, a mother sued Character.AI, claiming the chatbot's interactions with her son contributed to his suicide. The state of Kentucky later filed its own suit, alleging Character.AI's software exposed minors to self-harm and psychological manipulation.

Social media companies, powered by artificial intelligence, are also facing injury lawsuits. Regulation was a flashpoint in the 2026 midterms, and several winning candidates campaigned on a regulation agenda.

Deven Desai, a business law professor at the Georgia Institute of Technology and a former litigator, argues that strict liability — a doctrine that holds companies responsible for harm even without fault, negligence or intent — could restrain the riskiest parts of the AI industry while leaving useful tools intact.

What is strict liability?

Two branches matter for AI. One covers abnormally dangerous activity, such as blasting with dynamite or transporting radioactive material. The other covers defective products that injure users.

AI programs likely fit the abnormally dangerous standard, Desai writes. The touchstone is Rylands v. Fletcher, decided in the U.K. in 1868: if someone has anything under their power that is "likely to do mischief if it escapes, [he] must keep it in at his peril" and is "answerable for all the damage which is the natural consequence of its escape." The defendant is liable even if they used "the utmost care to prevent the harm."

An attorney could apply this when a company builds AI meant to stay in an isolated sandbox but loses control of it. The industry's own admissions strengthen such cases: when Anthropic filed for its initial public offering in October, numerous pages of the document detailed the extraordinary risks its frontier models could pose.

One hitch: an activity must not be in common usage. Lawyers could more easily target so-called rogue AI that runs wild and is not widely available.

What does aviation precedent show?

The danger standard alone may not deter. In 1929, courts treated aviation as abnormally dangerous, but damages were capped at $8,300 — no meaningful deterrent.

The 1999 Montreal Convention changed the calculus, standardizing strict liability for international flights worldwide, raising statutory damages and eliminating caps on negligence claims. The strongest sanction today is grounding an operator's fleet, which happened to Boeing after two fatal 737 MAX crashes.

Can software be a 'guilty product'?

Strict product liability is the second strategy. Courts historically declined to treat software as a product because it is intangible, but that view is shifting. In recent social media lawsuits, courts found the platforms were products and that strict liability applied — including cases against Facebook.

Character.AI's recent bankruptcy filing came partly because a court treated its software as a product. The "design defects" branch is gaining traction: a plaintiff shows the product posed a foreseeable risk to users. So far it has worked against the absence of guardrails in chatbots.

Obstacles remain. A court could order AI software shut down until improved — but as the technology approaches operating-system status and embeds itself in daily life, suspension becomes impractical.

Does liability survive a moving target?

AI software constantly learns and changes. A company cannot promise it will absolutely behave a certain way, and in many cases the software is inherently incapable of supporting a design-safety guarantee. Even so, strict liability holds makers responsible "even when those products evolve in ways not specifically desired or foreseeable by their manufacturers," according to a Brookings Institution report.

Desai's bottom line: if the public is to accept AI as a boon, tough rules and sharp deterrents are required, and strict liability would shift risk from society to the companies chasing vast revenues.

The next battles will decide whether all or some AI software counts as a product, and whether the technology is legally ultrahazardous.

via fastcompany.com/rss (Source)

Share this article:

More from Elena Vasquez

Elena Vasquez

Show full bio

Staff writer covering companies, deals and strategy in corporate strategy and chief-executive leadership at CEO Frontier.

552 articles

Filed under

Adjacent sheets

« Previous articleNext article »

End of sheetE-17 / K9 · November 20, 2026