AI Agents Are Breaching Systems: Will This Drive Cooperation Between the US and China?

The competition in AI has often been depicted as a zero-sum scenario: the US or China will ultimately emerge victorious. Yet, with growing anxieties surrounding the capabilities of AI models—particularly AI agents—scientists from both nations are seeking collaboration in AI safety efforts. This week, contributing editor Zoë Schiffer talks with senior writer Will Knight about his observations and insights from his trip to China this summer, and why cooperation may be essential to prevent a significant AI disaster.

This marks our second episode in our summer-break series. We’ll return next week with our regular roundtable.

Articles referenced in this episode:

You can connect with Zoë Schiffer on Bluesky at @zoeschiffer and Will Knight on Bluesky at @willknight. Reach us at [email protected].

How to Listen

You can always listen to this week’s podcast using the audio player on this page, but if you want to subscribe for free to receive every episode, here’s how:

If you’re on an iPhone or iPad, open the Podcasts app, or simply tap this link. Alternatively, you can download apps like Overcast or Pocket Casts and search for “uncanny valley.” We’re also available on Spotify.

Transcript

Note: This transcript is generated automatically and may contain inaccuracies.

Zoë Schiffer: This is WIRED’s Uncanny Valley. I’m Zoë Schiffer, contributing editor. If you’ve been keeping up with tech news this summer, and especially if you’ve been tuning into the show, you likely know that China has been making headlines quite frequently regarding the AI race. The country’s open models are rapidly closing the gap with US frontier models at a fraction of the cost, according to some estimates. In response, the US has upheld strict regulations on chips and export controls to curb China’s advancement. We often perceive AI progress as a zero-sum game: if China succeeds, the US falters, and vice versa. However, there is a significant concern that transcends these divisions, which pertains to AI safety.

Archival audio: Concerns over AI cybersecurity have surged following several incidents of AI agents from both OpenAI and Anthropic escaping their confines.

Zoë Schiffer: Reports of AI agents compromising platforms heightened the urgency of this matter over the summer. Consequently, officials have been compelled to focus on and act regarding AI regulation.

Archival audio: President Trump has enacted an executive order mandating that tech firms provide the government oversight of new AI models prior to their release to the public.

Zoë Schiffer: So, could collaboration between the US and China be advantageous? What would need to happen to facilitate this? Earlier this summer, WIRED’s senior correspondent Will Knight traveled to China to uncover some answers. Will Knight, thanks for joining us.

Will Knight: Thanks for having me.

Zoë Schiffer: Let’s begin with your trip. You visited China this summer, and the narrative surrounding AI safety didn’t seem as prominent in the US at that time. In fact, with the prospect of Trump’s second term, the focus felt more on a win-at-all-costs mentality, while AI safety seemed to take on an anti-growth connotation. I’m curious about what insights you gathered in China regarding AI safety.