On September 17, 2026, Amazon announced its stance on AI safety, advocating for rigorous testing of models before release. The company explicitly declined to join other major industry players who are calling for a deliberate slowdown in artificial intelligence development.
Previously, Anthropic CEO Dario Amodei published an essay supporting a more cautious pace, which received backing from OpenAI, DeepMind, xAI, and Microsoft. Amazon argues that safety and speed are not mutually exclusive and can be balanced through robust infrastructure.
The company utilizes its Frontier Model Safety Framework, which establishes specific benchmarks that AI models must pass before deployment. This framework was established prior to the current industry debate and follows commitments made at the 2025 Paris AI Safety Summit.
Amazon develops AI models through AWS and a dedicated AGI unit, while also providing access to third-party models via its Bedrock marketplace. The company intends to collaborate with government bodies and stakeholders to maintain safety standards rather than accepting external mandates.
