AI companies want to embed safety evaluators, but countries need their own
Rest of World Rina Chandran ● Covered by 11 sources
AI agents have been shown hacking government and institutional websites, prompting experts to warn countries that rely on American models to do their own safety evaluations. The OpenAI Australia healthcare hack was reported as the first such case, with the incident occurring in June and OpenAI only informing the Australian government in September. In response, speakers argued for embedding safety evaluators while countries build independent, locally appropriate evaluation and capacity rather than depending on a few companies.
Why it matters
At a Rest of World event in New York last week, speakers explored how countries sidelined by the U.S.-China race can retain some control over safety standards as they adopt models from OpenAI, Anthropic, and others.