The government wants AI companies to prove their tech is 100% safe before they release it, but it’s actually mathematically impossible to do that.
The study applies computer science theory to show that predicting whether a complex algorithm will cause 'harm' is a formally 'undecidable' problem, similar to the famous Halting Problem. This means the legal system is demanding a level of foresight that is physically impossible to achieve with any known form of computation.
Computational Impossibility of <i>Ex Ante</i> Compliance: A Due Process Analysis of Algorithmic Regulation in <i>NetChoice v. Bonta</i>
SSRN · 6209459
This paper analyzes the California Age-Appropriate Design Code Act (CA AADC) through the lens of computational complexity theory and procedural due process. While the litigation in NetChoice v. Bonta has primarily focused on First Amendment scrutiny, this analysis demonstrates a fundamental defect in the statute's mechanics: the requirement for ex ante semantic prediction of algorithmic outputs. By mandating that covered entities mitigate "harm" and prioritize "best interests" prior to the execu