China's leadership warns about AI risk, and Anthropic's mixed record

This issue covers how Chinese leaders are voicing concern about losing control of AI, why giving the public a financial stake in AI companies could weaken pressure to slow the race, and how AI is proving both risky and useful for children.

A report from Concordia AI finds that worries about losing control of AI have reached the top of China's leadership, with Xi Jinping and Premier Li Qiang both urging caution, though nearly all of the country's rules remain voluntary rather than binding (Robert)
Proposals from Gavin Newsom and others to give Americans a financial stake in AI companies show that public concern is shaping politics, but tying more people's money to the industry could reduce pressure for an international slowdown (Alana)
If language models stall, companies will pursue other approaches such as world models, which are trained to predict how an environment behaves rather than the next piece of text (Joe)
Anthropic says it is pushing state safety rules a little stronger each time, but its record includes weakening a California bill and dropping earlier safety pledges (Joe)
Common Sense Media rated Google's AI search an unacceptable risk to children after it missed warning signs in testing, though one of the two tools flagged risky behavior more often and some children have few better options (Joe)