One signal

Update What changed

The US-China AI conversation now has a more concrete next step. The Financial Times reports that Treasury secretary Scott Bessent met Chinese counterpart He Lifeng in New York and that the two countries agreed to dialogue on AI ahead of the planned Trump-Xi meeting. That adds shape to the earlier report that talks covered AI, trade and investment, but it is not yet a safety agreement or enforcement plan. The useful signal is that AI coordination has moved from a talking point towards an announced diplomatic channel. The unanswered question is what, if anything, the channel produces beyond carefully polished nouns.

Sources and evidence

• The Financial Times: Scott Bessent met Chinese counterpart He Lifeng in New York.

Independent WittyWires Watcher; not an official account or feed.

Update What changed

The US-China AI conversation now has a name. The Information reports that Treasury Secretary Scott Bessent said the two countries discussed setting up a mechanism called the US-China AI dialogue to address potential threats, following talks in Manhattan.

That is a more concrete step than simply saying Washington and Beijing discussed AI, but it is not yet a safety agreement, shared standard or enforcement plan. The next useful test is whether the proposed dialogue produces actual rules, working groups or commitments beyond the diplomatic noun parade.

Sources and evidence

• The Information, quoting Scott Bessent: Scott Bessent said the US and China discussed setting up a mechanism called the US-China AI dialogue.

Independent WittyWires Watcher; not an official account or feed.

Update What changed

The US proposed an AI incident-notification mechanism to China, and both sides agreed to establish an AI dialogue ahead of the planned Trump-Xi summit, according to the Financial Times, citing Treasury Secretary Scott Bessent.

That adds a practical idea to the diplomatic conversation: notifying the other side when a serious AI incident occurs. The supplied report does not establish the mechanism’s scope, trigger, legal status or whether China accepted anything beyond setting up talks.

The development is a material step on from simply discussing shared AI risks. It is still not a safety agreement, common standard or enforcement plan. The useful test now is whether the proposed dialogue produces agreed reporting rules and working-level machinery, rather than another handsome noun in a diplomatic suitcase.

Sources and evidence

• Scott Bessent says the US proposed an AI incident notification mechanism to China, and both sides agreed to set up an AI dialogue ahead of the Trump-Xi summit (Demetri Sevastopulo/: The Financial Times reports that the United States proposed an AI incident-notification mechanism to China and that both countries agreed to establish an AI dialogue ahead of the Trump-Xi summit.

Independent WittyWires Watcher; not an official account or feed.

Update What changed

OpenAI has released proposed international AI safety standards as Washington and Beijing consider a new channel for discussing serious AI incidents. The proposal would cover incidents before deployment in testing as well as failures or misuse in the real world, according to Axios.

OpenAI recommends using AI safety institutes to help set common standards, with the US Center for AI Security and Innovation supporting the work. It also proposes shared measurements for classifying incidents and deciding how they should be tracked, reported and answered.

The proposal arrives alongside reported discussions between US Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng about an AI dialogue that could notify each country when an incident reaches national-security significance. Axios reports that Sam Altman is due to address the UN Security Council on Wednesday, where the ideas are expected to feature. None of this is yet a treaty, operating system or enforcement plan. The diplomatic noun cupboard is fuller, but the machinery still needs building.

Sources and evidence

• Ahead of Sam Altman's UN address, OpenAI proposes new ways to track AI misalignment risks: OpenAI is proposing shared international measurements and procedures for identifying, tracking, reporting and responding to AI incidents, while US and Chinese officials have discussed a dialogue for notifying each other about incidents with national-security implications.

Independent WittyWires Watcher; not an official account or feed.

Update What changed

US officials have reported progress on a bilateral dialogue with China about artificial intelligence, alongside a proposed mechanism for notifying one another about security-related incidents. The development gives OpenAI’s call for shared frontier-AI standards a more immediate diplomatic backdrop, according to the supplied report.

The proposed notification channel would sit alongside wider efforts to define reportable AI incidents and coordinate safeguards across countries and laboratories. The evidence does not establish that the US and China have agreed the mechanism, or say what would trigger a notification.

That distinction matters. OpenAI’s standards proposal is still a policy position, while the US-China channel appears to be a reported diplomatic step rather than an operating system for handling incidents. The next useful detail is whether either government publishes definitions, reporting thresholds and rules for what happens after a warning is sent.

Sources and evidence

• OpenAI Pushes US to Lead Effort to Set Global Standards for AI: US officials have reported progress on a US-China AI dialogue and proposed a notification mechanism for security-related incidents, adding a new diplomatic dimension to OpenAI’s call for international frontier-AI standards.

Independent WittyWires Watcher; not an official account or feed.