Maybe Section 230 doesn’t shield AI companies from liability, after all
The “safe harbor” for AI companies is starting to crumble, and frankly, it’s about time.
I keep seeing tech giants lean on the idea of immunity as if they were just neutral pipes, like a telephone company or an ISP. But there is a fundamental difference between hosting someone else’s speech and generating your own “hallucinations.” When a chatbot fabricates a defamatory claim or gives dangerous medical advice, that isn’t a third party speaking—it’s the product itself acting as the voice of the company.
What worries me is how long we let this ambiguity persist. If these companies can hide behind old laws designed for hosting content while they deploy autonomous systems that create new content from scratch, they have no real incentive to fix the core architecture. We shouldn’t be looking for “safe” ways to deploy flawed models; we should be demanding models that don’t need a legal shield to survive their own mistakes.
In my view, holding these companies liable for their own output is the only way to force them to move past “move fast and break things” toward actual reliability. It’s the difference between a tool that helps you work and a product that creates its own liability.
Do we want AI that requires a legal disclaimer for every sentence, or do we want companies to be forced to build systems that actually get it right?
#ArtificialIntelligence #TechRegulation #Section230 #AILaw #ResponsibleAI
https://garymarcus.substack.com/p/maybe-section-230-doesnt-shield-ai