Amazon calls for rigorous AI testing, avoids backing slower development

Amazon has weighed in on one of the most contentious debates in tech: how fast should AI development move? The company’s answer, essentially, is “as fast as you want, as long as you’ve done the homework.” On September 17, 2026, Amazon publicly staked out its position on AI safety, arguing that models should only be released after rigorous testing proves they’re safe and reliable. It’s the company’s first substantial comment on the existential risks of artificial intelligence, and it lands with a notable twist: Amazon is not joining the growing coalition of industry heavyweights calling for a deliberate slowdown. A different flavor of caution About a week earlier, Anthropic CEO Dario Amodei published a detailed essay advocating for a more cautious pace of AI development. OpenAI, DeepMind, xAI, and Microsoft all signaled support for Amodei’s slower-is-safer philosophy. Instead of endorsing deliberate deceleration, Amazon emphasized a framework built around testing thresholds and collaborative safeguards, arguing that safety and speed aren’t inherently at odds. This isn’t just theoretical posturing. Amazon already operates what it calls a Frontier Model Safety Framework, a set of spec...