Anthropic与OpenAI推动AI安全测试，引发利益冲突质疑

Anthropic and OpenAI are pressing for stronger testing of advanced artificial intelligence, drawing scrutiny over how much influence the companies should have over the rules governing their own technology. Both companies argue that increasingly capable systems require deeper independent assessment and more consistent safety standards. Their proposals would give outside specialists greater access to development processes that have largely remained under corporate control. But a September 27 Associated Press report highlighted concerns that the same initiatives could strengthen the leading laboratories’ commercial positions. The debate now extends beyond whether AI needs testing to who selects the evaluators, determines their access and decides what the public gets to see. …