Amazon Says AI Models Should Be Released When ‘Ready and Safe’

(Bloomberg) -- Amazon.com Inc. said artificial intelligence models should be released only after rigorous testing, weighing in on the debate about the pace of AI development after top labs called for a slowdown following safety lapses.

"We don't see it as a choice between progress and safety," Amazon spokesperson Tom Parnell said in an email. "Models should be released when they're ready and safe to use, which comes from rigorous testing and strong safeguards. There are risks if that doesn't happen collectively, and we believe that as an industry, in partnership with government, we'll work through the right protections."

Amazon is partner of and investor in leading AI companies, including OpenAI and Anthropic PBC, whose executives have called for moderating the pace of cutting-edge AI work to assess the safety of the technology.

Amazon develops its own large language models, though they are not seen as rivals to the most powerful tools from OpenAI, Anthropic and top Chinese companies. The company's Amazon Web Services unit also sells computing power and infrastructure management tools to the AI industry.

"AI is already delivering real value to customers, so getting this right matters," Parnell said. Reuters reported the company's comments earlier.