Table 4.

Relative benefits of different scopes of cooperation to avoid harms specific to AI types, split across development and deployment phases of the AI life cycle (+ +  = very useful, +  = somewhat useful, − = not useful)

	Unilateral regulation of foreign AI		Selective cooperation		Global cooperation
	Development	Deployment	Development	Deployment	Development	Deployment
Foundation models	−	+	+	−	+  +	−
AI-powered physical products	−	+	+	−	+  +	−
Small-scale software as a service	−	+	+	−	+  +	−
Militarily relevant AI	−	−	+	+	+  +	+  +