Episode Summary
Executive Summary: The episode argues that AI should be regulated primarily through liability and duty-of-care rules, not just narrow risk-based laws. Casey Mock explains that treating AI like a manufactured product would push companies to build safer systems, warn users, and empower internal legal teams to deter harmful deployment, while still preserving innovation and allowing courts and regulators to adapt over time.
Main Topics: Why liability is the right framework for AI (Priority: 5/5): The conversation centers on applying products-liability/negligence principles to AI so companies can be held accountable when their systems cause foreseeable harm, similar to other manufactured products. AI as a product, not just a service or code (Priority: 5/5): Casey argues that AI systems should be treated as complex manufactured products even if they are intangible, because the legal purpose is consumer protection and accountability, not tangibility. Concrete harms and use cases (Priority: 5/5): The hosts discuss deepfake abuse, scam calls/videos, and chatbot misinformation as examples where victims are often unable to identify the actual wrongdoer, making developer liability more effective. Balancing accountability with innovation (Priority: 4/5): The framework is presented as pro-innovation and pro-business because it encourages safer product design without prescribing one technical solution, and avoids overburdensome pre-clearance regimes. Enforcement, private lawsuits, and government action (Priority: 4/5): The proposal combines a limited private right of action with government enforcement so victims are protected without being forced to relive trauma, and mass actions can shape industry behavior. Comparison to SB 1047 and other tech regulation (Priority: 4/5): The episode contrasts CHT’s approach with California’s SB 1047, arguing that risk-specific laws can be weaker if they work backward from catastrophic harms rather than establishing a general duty of care. Political feasibility and global impact (Priority: 3/5): The speakers suggest the framework has bipartisan appeal, could influence how firms behave internally, and may shape AI deployment standards globally because of the U.S. market’s size.
Key Arguments: Current AI incentives reward speed and flexibility, not safety; liability would realign incentives toward responsible design. Treating AI like a product under established negligence/products-liability principles is more legally coherent than leaving harms to victims alone. Courts need legislative clarification because AI is evolving too quickly for common-law adaptation to happen slowly and organically. A duty-of-care framework can be technology-neutral and not over-prescribe technical fixes, allowing companies to innovate on safety. Real-world harms like deepfake abuse and chatbot misinformation often leave victims unable to sue the true perpetrator, so developer accountability is necessary. Government enforcement should complement private lawsuits because some harms are traumatic or diffuse and are better addressed by attorneys general or agencies. Fears of frivolous litigation and innovation chill are described as standard industry arguments that historically have not prevented safer, better products. By making safety a business and legal concern rather than a PR concern, companies’ internal attorneys and safety teams gain more leverage over engineering and deployment decisions. A liability regime could create market differentiation, with companies competing on safety in the way car manufacturers compete on safety. The framework is intentionally a first step: it addresses direct injuries now and creates a foundation for broader AI regulation later.
Data Points: Years of tech policy experience: about 10 years - Casey Mock describes his background in tech policy. Historical duration of products liability principle: about 110 years - Casey notes that negligence/products-liability principles have long applied in American commerce. Canadian airline chatbot case: 1 customer harmed by incorrect bereavement policy guidance - Example of a chatbot giving wrong advice that the airline allegedly refused to honor. Hong Kong deepfake scam loss: $25 million - Example of a finance worker deceived by a deepfake video impersonating a CFO. Antitrust and safety team budget example: Facebook civic integrity funding came from antitrust budget - Sasha references Frances Haugen’s account of how companies justify safety spending. Ford Pinto recall calculus: roughly a decade - Used as an analogy for companies comparing settlement costs with recalls. National Traffic and Motor Safety Act example: 1966 - Historical analogy for industry claiming safety regulation would harm innovation. Child labor ban passage: 1937 - Cited as another case where industry warned regulation would bankrupt business. CBH white paper filing requirements: light-touch / minimal - Casey describes the compliance burden as modest rather than FDA-style pre-clearance.
Pivotal Quotes: "What artificial intelligence systems are and what social media systems are in principle is that they are very complex manufactured products and they can cause real harm out in the world and they can hurt people." — Casey Mock: Argument for treating AI under products-liability-style principles. "The specter of a lawsuit will empower the attorneys to go into these meeting rooms with business teams and wag their fingers and say, No, you can't do that anymore." — Casey Mock: Explaining how liability changes internal company incentives before litigation even happens. "Let's not have an optical lipstick of proving with PR that we're doing things for safety. Let's talk about meaningful changes that will actually deter the worst things from happening." — Tristan Harris: Critique of superficial safety efforts and support for enforceable accountability.
Implications: If adopted, this framework could force AI firms to build safer products, strengthen internal safety/legal checks, and make responsible deployment the default. It may also set a U.S. standard that influences global AI practice.