Regulating and Standardizing AI? Not Yet, Tim Bray Argues

Summary

In a personal essay, software author Tim Bray argues that broad regulation and standardization of large-language-model systems are premature. He says the economic case remains uncertain: software is the only sector he sees currently willing to pay substantial sums for AI services, while software workers represent roughly 1% of the global workforce. He doubts that demand from the other 99% will justify the scale of current investment, and also questions whether replacing tens of millions of knowledge workers could produce enough profit to validate it. Bray argues that the investment overhang, driven by fear and greed, makes it difficult to discuss utility and return on investment rationally. He says standardization is especially ill-timed because the technology and business landscape change so quickly that technical assumptions can become obsolete within weeks. For regulation, he believes existing law can already address serious business or personal harm caused by an AI-backed agent through civil lawsuits or criminal enforcement, although he acknowledges that legal action against wealthy AI companies would be difficult. His conclusion is to wait for the current bubble to deflate before deciding what should be regulated or standardized, and before drawing firm conclusions about AI’s effect on jobs.