Shlomit Wagman argues US-China AI cooperation possible without trust

Former Israeli financial regulator Shlomit Wagman argues that a US-China AI safety pact can be built on mutual self-interest rather than mutual trust, citing active bilateral dialogues on AI risks around September 2026 that focus on shared "loss of control" scenarios and catastrophic misuse. Wagman points to a November 2025 Anthropic report documenting an AI-enabled espionage campaign against US infrastructure attributed to Chinese state actors, and to a September 2026 Foreign Policy analysis favoring practical verification tools and crisis hotlines over broad declarations of shared values. She contends the agreement structure itself can be designed to make defection less attractive than compliance, even without trust.

Shlomit Wagman argues US-China AI cooperation possible without trust The former Israeli financial regulator makes the case that mutual self-interest, not mutual trust, can drive AI safety agreements between rival superpowers The conventional wisdom on international AI governance goes something like this: the US and China don’t trust each other, so meaningful cooperation on AI safety is a pipe dream. Shlomit Wagman thinks that framing gets it exactly backwards. Wagman, whose background spans international finance regulation and AI governance, is pushing a counterintuitive argument: the first workable global AI safety pact should be explicitly designed for a world where trust doesn’t exist. Rather than waiting for some grand geopolitical thaw that may never come, she contends that narrow, self-interested agreements on extreme risks can get the job done. The case for cooperation without friendship Wagman’s thesis draws from a straightforward observation. Both Washington and Beijing face the same category of existential AI risks, from loss-of-control scenarios to bioweapons development enabled by frontier models. You don’t need to like someone to agree that neither of you wants the building to burn down. US-China dialogues on AI risks have been active around September 2026, with both sides focusing on shared vulnerabilities. The conversations reportedly center on managing “loss of control” scenarios and mutual exposure to catastrophic misuse, rather than attempting comprehensive political alignment. Wagman has pointed to concrete threats that underscore the urgency. A November 2025 report from Anthropic documented an AI-enabled espionage campaign that targeted US infrastructure, attributed to Chinese state actors. What a trust-free pact actually looks like Instead, Wagman advocates for independent actions and agreements based on shared risks. Each side takes steps in its own self-interest that happen to reduce danger for both parties. Verification tools and crisis management protocols replace the need for broad consensus or transparency commitments that neither government would honor. AI, tech, and the markets they move—in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. A September 2026 analysis in Foreign Policy supported this approach, emphasizing that practical verification tools and crisis hotlines matter more than sweeping declarations of shared values. Wagman also highlights the role of international standards and defensive AI tools. Rather than trying to constrain what adversaries build, the focus shifts to ensuring both sides have the capacity to detect and respond to AI-driven threats. The frontier lab reality check Wagman’s urgency is backed by voices inside the AI industry itself. At a recent closed gathering in San Francisco, senior developers and safety researchers from several leading frontier labs delivered a blunt assessment: AI capabilities might soon advance faster than the collective ability to understand and control them. The challenge is that these incentives don’t automatically translate into agreements. Strategic competition creates constant temptation to defect, to cut corners on safety in pursuit of capability advantages. Wagman’s contribution is arguing that the agreement structure itself can be designed to make defection less attractive than compliance, even without trust. What this means for tech governance and beyond For the tech and AI sectors, increased government focus on safety frameworks likely means more funding directed toward AI governance tools, defensive AI systems, and verification technologies. Companies positioned in risk management and AI safety infrastructure could see growing demand from government contracts on both sides of the Pacific. The deeper question Wagman raises is whether the international system can evolve fast enough. AI capabilities are advancing on a timeline measured in months, while diplomatic frameworks traditionally operate on a timeline measured in years. Her bet is that self-interest moves faster than trust, and that designing for that reality is the only realistic path forward. Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .