Xi flags AI "loss of control" as China's open models draw scrutiny

China's president raised the risk of losing control over AI at a Shanghai conference, a new Chinese open weight model prompted overstated claims that China has caught up to the United States, and a Politico survey found more agreement inside the AI safety debate than expected.

Xi Jinping used the phrase "loss of control" and said AI must stay under human control in his first keynote at the World Artificial Intelligence Conference, though he offered no compute thresholds, evaluations, red lines, or verification methods. (Donald)
Because any scenario where humans lose control of AI is also one where the Chinese government loses control, the speech suggests China may be willing to negotiate, which weakens the usual objection that no treaty is possible. (Donald)
Moonshot AI's Kimi K3 is the largest model whose weights anyone can download, but coverage repeated the company's benchmark and hardware claims without noting that benchmarks are noisy and easy to game. (Joe)
The UK AI Security Institute places the American lead at four to seven months, and once weights are public the release cannot be undone and safety protections can be removed by anyone who downloads them. (Joe)
Politico's map of AI safety factions shows four of seven groups worry about extinction risk and most want AI kept under human stewardship, though they disagree on whether guardrails suffice or development should pause. (Alana)