The United States and nine Pax Silica partners signed the San Francisco Declaration on November 15, 2026, setting out cooperation on protecting the technologies and infrastructure behind advanced AI. The State Department’s statement describes a shared security agenda with implications for research computing, international technology partnerships and university research-security planning.
What the partners signed
The State Department media note, distributed by Public Technologies, identifies the signatories as the United States, Australia, India, Israel, Japan, South Korea, Qatar, Singapore, the United Arab Emirates and the United Kingdom. It places the signing at the Pax Silica Founders Retreat convened by Under Secretary of State for Economic Affairs Jacob Helberg.
The statement proposes five working groups covering innovation policy, adoption in key sectors, cooperation across the technology supply chain, safe harbors and security hubs, and security and resilience. Its definitions encompass sensitive technology and the physical and digital infrastructure supporting it. A footnote explains that some participants use artificial intelligence for the declaration’s term super intelligence.
Why the agreement reaches beyond AI chips
KBS World reported on November 16 that the initiative connects semiconductor design, manufacturing equipment and the ports used to move products. Its account presents the agreement as part of Washington’s effort to reduce dependence on China across the AI supply chain, including critical minerals.
That framing matters for university computing discussions: the relevant exposure is broader than the model a laboratory chooses to run. CASRAI’s editorial interpretation is that a useful institutional assessment would connect computing procurement with infrastructure continuity and international collaboration, rather than assign the story only to an AI software team.
The wider industrial emphasis also appears in Helberg’s prepared congressional testimony. That document describes Pax Silica in terms of logistics, industrial production and AI infrastructure, including data centers and connectivity. It discusses reducing strategic dependency and building redundancy, and links technology protection with industrial incentives and AI adoption. These are the administration’s stated objectives, rather than evidence that particular projects have already delivered results.
European concerns predate the signing
The November 15 declaration followed disagreement over its scope. Euronews reported on November 12, after seeing a draft, that EU governments were concerned about broad definitions, regulatory autonomy and the terminology surrounding superintelligence. The outlet said some member states wanted objective criteria for sensitive technologies and critical infrastructure.
Euronews also reported that the European Commission had received a mandate to negotiate, while authorization to sign in time for the ceremony remained uncertain. That distinction helps explain why participation in the broader Pax Silica initiative and signature of this particular declaration should be treated separately. The report described European concerns about Washington’s discretion and the potential pressure on partners to follow its approach.
For cross-border academic partnerships, the editorial implication is to identify the actual instrument and jurisdiction being discussed. A coalition’s general direction and an institution’s present obligations answer different questions. Existing arrangements should be assessed against the rules and contracts that govern them, with the new declaration tracked as a development in the policy environment.
What research offices can take from the announcement
The declaration brings questions familiar to research security into a wider conversation about AI infrastructure. Our interpretation is that institutions should be ready to connect three practical discussions: which external infrastructure research depends on, which partners supply it, and who follows changes affecting access. Those questions can help research, procurement and computing teams discuss the same dependencies.
For the research security officer, a useful next step is to distinguish an announced diplomatic commitment from an applicable institutional requirement. For computing teams, the corresponding task is to document the dependencies that would matter if a supplier or partnership changed. Neither exercise requires forecasting a chip shortage or assuming that a particular collaboration will be restricted.
What happens next
Anadolu reported on November 16 that participants would establish working groups on policy, adoption, cooperation and security. Citing Helberg’s comments to the Financial Times, it also reported that Germany, Sweden, the Netherlands and the European Commission were expected to sign the following week. That remains an expectation, rather than confirmation of their signatures.
The developments to watch are those working groups’ outputs and any subsequent national measures. Institutions following AI export controls will need to examine the terms of any later action before drawing conclusions about a specific research project, computing service or international partnership.
Sources
- U.S. Department of State, San Francisco Declaration media note and text, November 15, 2026, distributed by Public Technologies. Checked 2026-11-16.
- KBS World, AI technology and supply-chain accord report, November 16, 2026. Checked 2026-11-16.
- U.S. House of Representatives, prepared testimony of Jacob Helberg, background document. Checked 2026-11-16.
- Euronews, Luca Bertuzzi, European concerns about the draft declaration, November 12, 2026. Checked 2026-11-16.
- Anadolu, Seyit Kurt, U.S. and nine partners sign AI supply-chain agreement, November 16, 2026. Checked 2026-11-16.







