Amazon urges rigorous testing for AI models, stops short of slowing development

Amazon entered the debate over artificial-intelligence safety, stating that AI systems should be deployed only after comprehensive testing and robust safeguards are in place. While acknowledging risks, the Seattle-based firm stopped short of supporting the recent push by Anthropic, OpenAI, DeepMind and others for a slower development pace. Amazon develops its own models via Amazon Web Services and an AGI division that powers Alexa, and it also provides infrastructure and the Bedrock marketplace for other labs.

The remarks were made after Anthropic’s chief Dario Amodei published a lengthy essay urging industry pacing, and after Nvidia and Meta CEOs distanced themselves from broader responsibility claims. The company did not disclose whether it will adopt new safety protocols, and it remains unclear if it will be invited to upcoming White House talks with AI executives, including President Donald Trump’s scheduled meeting.