Seth Center urges testing China's AI safety cooperation ahead of Trump-Xi talks
Source: Fox Business · All Fox Business reports
Unlock the full scoreboard
Letter grade, factuality, lean, and rationales — free with registration. No card required.
See grades free How grading works
Already have an account? Sign in. The full report below is free to read.
Embed this grade
Disagree with this grade or political lean?
Flagging is open to every reader with a free account. Sign in or create one to dispute this report.
Topics in this report
Summary
The segment discusses upcoming talks between President Trump and Xi Jinping, focusing on a proposed US-China AI 'hotline' or notification mechanism for national-security incidents. Host Brian and Dagen introduce the topic by noting US AI leadership and Trump's commitment to not hindering growth, then interview former State Department official Seth Center, who expresses skepticism about comprehensive cooperation. Center argues for testing China's willingness on narrow AI safety issues like cybersecurity and biosecurity benchmarks while warning of propaganda efforts to exacerbate US domestic debates over data centers. The conversation touches on unrealistic expectations for treaties, Chinese open-weight models as a potential threat, and the value of baseline discussions.
Sourcing relies on Center's firsthand experience from prior dialogues and his recent op-ed, with references to Treasury Secretary Bessent's push for transparency. No anonymous sources or graphics; the throughline is pragmatic caution—engage to test sincerity but prioritize domestic safeguards and verification over broad agreements.
Editorial Assessment
The broadcast accurately conveys Center's core arguments from his New York Times piece that China prioritizes competitive advantage in AI talks, with limited interest in genuine safety cooperation beyond gaining edges like exploiting US models. Claims about Chinese propaganda amplifying US data-center opposition are well-supported by OpenAI and X reports on influence operations using AI-generated content to stoke local concerns over energy costs and infrastructure. Viewer perception may skew toward viewing all cooperation as futile due to the 'fool's errand' framing by hosts, missing nuance that Center sees value in testing propositions and building verification mechanisms. Omitted context includes China's advancing domestic AI safety benchmarks, regulatory focus on agentic risks, and recent proposals for tiered governance of open-weight models, which complicate the purely adversarial picture. Overall quality is solid for a short cable segment—claims hold up against primary sources—but selective emphasis on threats risks overstating hopelessness while underplaying mutual interests in incident notification ahead of the Trump-Xi summit.
Key Moments
China has little interest in AI safety talks beyond gaining competitive advantage and views US risk concerns as disadvantageous
Directly matches Center's NYT op-ed recounting first US-China AI dialogue where officials resisted safety details and pushed for process while seeking concessions on chips.
US should test if China will carve out AI safety from geopolitical competition, pressing on why Chinese models lag on cybersecurity and biosecurity benchmarks
Center's explicit recommendation; 2026 reports confirm gaps in Chinese model performance on some misuse benchmarks like CyBench and biosecurity refusal rates, though China is developing its own safety evaluations.
China uses propaganda, including AI-generated, to inflame US domestic debates over data centers and energy costs to slow American AI buildout
Supported by OpenAI's June 2026 report on 'Data Center Bandwagon' campaign, X disclosures of 200k suspected Chinese accounts, and analyses by Alethea and congressional testimony.
Comprehensive AI treaty with China is unrealistic; baseline discussions are not particularly harmful but expectations must be low
Center's position in the interview and op-ed; aligns with expert consensus that talks may yield only modest, nonbinding commitments or incident-notification channels like the proposed Bessent-He hotline.
Open-weight models from China pose real threats as they can be manipulated for nefarious purposes with no effective oversight
Valid risk per CSIS and CNAS reports on proliferation and distillation attacks, but omits China's emerging tiered governance signals, voluntary benchmarks, and that both US and Chinese models face similar misuse challenges.
Notable Concerns
- Hosts' introductory framing casts cooperation as largely pointless before Center offers a more measured 'test the proposition' view
- Limited exploration of China's documented progress on AI safety standards and benchmarks, which could provide counter-evidence to total disinterest
- Ends abruptly on open-weight risks without detailing US countermeasures or shared vulnerabilities
Sources Consulted
- Opinion | China Cares More About A.I. Victory Than Safety - The New York Times
- ‘The scientific, technical term is BS’: Inside the new Bessent-China AI hotline - POLITICO
- U.S. proposes emergency AI hotline with China - Axios
- China, Russia and Others Seek to Inflame Debate Over A.I. Data Centers - The New York Times
- “Data Center Bandwagon” Campaign: US-targeted influence activity - OpenAI
- State of AI Safety in China (2026) - Concordia AI
- China’s Open-Weight Challenge to U.S. AI Leadership - CSIS
- Red Lines: Understanding the National Security Risks of China’s Advanced AI - CNAS