The Biggest Questions About AI
The map · 4 Power · 4.4 International coordination and verification · 4.4.1

Common interests

Which risks are sufficiently shared that rival states could cooperate despite broader geopolitical conflict?

The candidate zones: evaluation science, verification technology, and loss-of-control red lines — areas where rivals share downside the way they shared it on nuclear accidents. Bengio and others have mapped the technical overlap.

View on the map → · Open in Browse →

What changed
November 2025–August 2026 · swept August 3, 2026 · editorial review pending
01

The two-year freeze ended: the US and China announced an intergovernmental dialogue on AI safety, and April's IDAIS-London statement shows where rival states' scientists still co-sign — near-term AI cyber capability handing non-state actors state-level attack capacity, with biological misuse close behind. The design question is now which cooperation format survives rivalry: standing information-sharing exchanges rather than arms-control bargaining, even as the one existing military commitment (human control of nuclear decisions) loses definition.

Recent thinking
4 featured from 7 tracked · November 2025–August 2026 · all 7 chronologically →
Concordia AI · 14 Jul 2026 report
State of AI Safety in China (2026)
China and the US announced an intergovernmental AI dialogue, ending a two-year freeze in official bilateral engagement on AI safety.

The definitive annual mapping of China's AI safety ecosystem, documenting the resumed US-China intergovernmental dialogue and China's UN and WAICO moves — the evidence base for which risks Beijing treats as shared.

International Dialogues on AI Safety participants · IDAIS (Royal Society) · Apr 2026 report
IDAIS-London Consensus Statement
Societies remain unprepared for this imminent threat, and biological misuse risks are not far behind.

The April 2026 track-II dialogue where senior US, Chinese, and other scientists jointly warned that near-term AI cyber capabilities will hand non-state actors state-level attack capacity — the red-line zone where rivals' scientists still co-sign.

Mark MacCarthy & Carl Schonander · Brookings · 14 Jul 2026 essay
How the US and China can cooperate to reduce urgent AI risks
it should be a regular process of sharing information, ideas, plans, and research in the area of AI model risk assessment and mitigation.

Design proposal for the new bilateral channel: standing technical exchanges on model risk assessment modeled on information-sharing rather than arms-control bargaining — an argument about which cooperation format survives geopolitical rivalry.

IISS · International Institute for Strategic Studies · Jun 2026 essay
Military AI governance under strain: the US–China dialogue
the single military-AI commitment that the two powers already share – keeping humans in control of nuclear decisions – loses definition.

Sober assessment of the military side of the shared-risk zone: the one existing US-China commitment is eroding in meaning just as the broader dialogue opens.

Additional relevant discussion (3)
AI Safety in China #23 — Concordia AI · AI Safety in China (Substack) · 10 Nov 2025
How the US and China Can Cooperate on AI Biorisk — Nick Corvino · ChinaTalk (Jordan Schneider) · 8 Sep 2026
How Trump and Xi Can Do AI Safety — Jay Kimmel · ChinaTalk (Jordan Schneider) · 4 Sep 2026
Foundational reading (2)In Which Areas of Technical AI Safety Could Geopolitical Rivals Cooperate?Bucknall, Trager, Bengio et al. · 2025Can the US and China Cooperate on AI?Brookings · 2025
Next4.4.2 Verification