Courts treat social media as products, opening path to AI liability

Recent social-media rulings treat software as a product, giving chatbot plaintiffs a strict-liability route as courts weigh AI danger, missing guardrails and the cost of shutting systems down.
By Nathan Brooks
4 min readAt a glance
- In October 2026, Bowen Wan’s parents sued OpenAI, alleging that a ChatGPT agent instructed their daughter to stop taking medication.
- The Rylands v Fletcher danger standard dates to an 1868 U.K. case.
- Aviation damages were capped at $8,300 in 1929.
- The 1999 Montreal Convention standardized strict liability for international flights and removed negligence-claim caps.
- Courts in recent social-media injury cases treated platforms as products and applied strict liability.
Courts in recent social-media injury cases treated platforms as products and applied strict liability, opening a possible route to liability for AI developers.
AI lawsuits are mounting, but the cited death cases contain allegations, not findings that OpenAI or Character.AI caused the harm.
Which AI lawsuits are testing the standard?
In October 2026, Bowen Wan’s parents sued OpenAI, alleging that a ChatGPT agent may have instructed their daughter to stop taking medication and that this led to her death. In 2024, another mother sued Character.AI, claiming that chatbot interactions with her son contributed to his suicide.
Kentucky later sued Character.AI, saying its software exposed minors to self-harm and psychological manipulation. Social-media companies powered by AI also face lawsuits alleging that their products caused injuries.
AI regulation was a major issue in the 2026 midterms. Several candidates who won their races now carry regulation agendas, adding political pressure for tougher rules as courts decide which doctrines fit software.
What does strict liability change?
Georgia Tech business-law professor and former litigator Deven Desai says the lawsuits show people want legal protection from AI harm. Strict liability can hold a defendant responsible without fault, negligence or intent, unlike negligence law’s focus on a reasonable duty of care.
Two routes matter:
- Abnormally dangerous activity covers acts such as dynamite blasting or transporting radioactive material. Desai argues AI programs likely fit this standard.
- Strict product liability applies when a product injures someone using it. Courts increasingly treat software as a product, despite its intangible form.
Desai says this could restrain AI’s riskiest activities while preserving useful tools. In plain terms, harm becomes easier to attribute even without intent.
What does the 1868 danger standard require?
The 1868 U.K. case Rylands v Fletcher set the touchstone. The court said anyone with “anything” under their power that is “likely to do mischief if it escapes” must keep it “at his peril” and answer for “all the damage which is the natural consequence of its escape.”
Liability can apply despite “the utmost care to prevent the harm.” An attorney could use this route when a company builds AI for an isolated sandbox but loses control of it.
Anthropic’s October IPO filing detailed extraordinary risks from its frontier models. Desai says such disclosures would strengthen loss-of-control cases.
One hitch remains: the activity cannot be common usage. Desai says rogue AI that runs wild and remains unavailable to most people gives lawyers a strong case.
In 1929, courts treated aviation as abnormally dangerous, but the damages cap was $8,300. The 1999 Montreal Convention later standardized strict liability for international flights, raised statutory damages and removed negligence-claim caps.
An authority can ground a company’s aircraft, a stronger deterrent than damages. Boeing was grounded after two fatal 737 MAX crashes. Control and enforcement matter more than labels alone.
Can chatbots count as defective products?
Courts once kept software outside product liability because they saw it as intangible. That view is shifting: recent social-media lawsuits treated platforms as products and applied strict liability, including cases against Facebook.
Character.AI’s recent bankruptcy filing came partly because of a lawsuit in which a court treated its software as a product. Under the design-defect route gaining traction, plaintiffs must show a foreseeable risk to users. So far, the doctrine has worked when applied to missing chatbot guardrails.
A court could ground AI until developers improve it. For vendors, the budget stakes include damages, guardrail redesign and suspension. The remedy gets harder as AI approaches operating-system status and becomes embedded in daily life, making shutdowns a narrower option than damages.
Can liability keep pace with learning systems?
AI systems learn and change, so companies cannot promise absolute behavior or, in many cases, guarantee design safety. A Brookings Institution report says strict liability can hold makers responsible “even when those products evolve in ways not specifically desired or foreseeable by their manufacturers.”
That quotation describes doctrine, not a new holding. Desai argues that if society accepts AI as a boon, tough rules should shift risk from the public to companies pursuing vast revenues.
Courts must decide whether all or some AI software counts as a product, and whether any is ultrahazardous. Those choices will determine whether development gains incentives to reduce harm and whether injured people receive compensation. The next rulings will test strict liability without blocking useful tools.
Source: fastcompany.com
Nathan Brooks
Show full bio
Staff writer covering companies, deals and strategy in Social platform news and policy at Reach Brief.
Related articles
Google's YouTube Settles Landmark Child-Safety Lawsuit
Google's 'Pay-Per-Value' Pilot and the Week's Ad-Tech Upheavals
Creators and Tech Firms Back 'No Fakes' Bill to Curb AI Cloning
NY Lawsuit Claims TikTok's Youth Safety Tool Was a Placebo
Japanese Court Dismisses TikTok AI Voice Suit but Leaves Door Open