US and China propose hotline to de-escalate AI threats ahead of Washington summit

The U.S. and China have proposed developing a hotline for AI incident reports to address the risk of a rogue AI model getting out of control and to prevent misinterpretations of attacks that could trigger rapid autonomous escalation, Reuters reports. This news comes ahead of a planned visit by Chinese Premier Xi Jinping to Washington on September 24. U.S. Treasury Secretary Scott Bessent has also confirmed that follow-up AI safety discussions are scheduled between the two nations in two months' time.

All of this sits against the backdrop of ongoing U.S. China trade and relations talks, in addition to various frontier AI agents seemingly breaking out of their containment and hacking organizations without oversight from humans.

While some are raising the alarm about the humanitarian risk posed by AI models, the U.S. and China are taking steps to ensure that if either party does end up triggering a dangerous AI event, there may be the opportunity for a joint response from both nations ahead of any potential perceived threat or escalation.

As artificial intelligence continues to make headlines circling around safety, both the U.S. and China now appear to be building global safeguards through a dedicated hotline to discuss any AI-related issues that escalate to national security threats.

Keeping humans in the loop

Following discussions between Scott Bessent and Chinese Vice Premier He Lifeng in New York over the weekend, the two sides have proposed that President Trump and Xi Jinping discuss various AI safety mechanisms, including a notification system focused around national security. This could include warnings of detected rogue AI actors operating anywhere in the world, as well as informing the other partner if an autonomous system performed an action that could be misinterpreted as an attack.

The scenario both parties want to avoid is an AI-driven autonomous weapon system - be they kinetic or something more asymmetric in nature, like AI-driven cyber attack capabilities - responding to provocation without consideration by a human. That could then trigger an escalated response in another AI system, leading to rapid escalation to full-blown conflict at a pace where humans wouldn't be able to intervene.

The proposed system could also help both parties develop joint responses to positive goals, as well as potential threats. This could form the basis for joint development pacing, considering the recent calls from U.S. AI firms to slow AI development to improve alignment and safeguards. China is less keen, suggesting that such a move could be used to hamper Chinese AI developments.

It's not clear at this time how keen China is to implement the proposed notification system, but analysts and experts on both sides have called for a system like this to be developed. Both the U.S. and China have agreed to further broach AI safety concerns in the coming months.

Laying the groundwork

We're still in the very early days of these discussions, with only the bare bones of a proposal for a hotline in place, and much to discuss on alignment between the two major AI-developing countries. The next step will be the talks between the two leaders, where much more than AI will be discussed.

If the relationship between the U.S. and China can move beyond its ongoing rivalry, the potential is there for much firmer agreements on the specifics around AI integration in national security and military actions. Although some analysts remain sceptical of firm agreements because of the limited trust between the two parties, there is some hope that agreements can be made on the direct weaponization of AI, core principles of AI safeguarding, and preventing them from targeting or damaging critical infrastructure.

The hope for a successful meeting is very much bipartisan. Leading Democrat in the U.S. House, Representative Ro Khanna, said that the first thing President Trump and Xi Jinping should work on in their meeting is developing core AI principles. Particularly around banning recursive self-improving AI and removing the option to use such autonomous systems on biological and nuclear weapons technology.

Washington's summit between President Trump and Xi Jinping is scheduled to begin in just a few days, where much more will be discussed beyond AI safety, as the rest of the world waits with bated breath, as the meeting between the two leaders could set the stage for another round of discussions surrounding the two nations' economic relations.