Staying in control once data moves
Introduction
Organizations have spent years hardening the places where confidential information lives, while the information itself rarely stays put. Contracts travel to outside counsel, client files move to auditors and service providers, and AI assistants pull documents into prompts that may be processed far from where the files are stored. Across the security industry, the focus is shifting accordingly, toward maintaining control over how information moves, who accesses it and what happens to it once it leaves the organization.
That shift reframes data sovereignty, exposes secure data exchange as an underestimated challenge, and raises new governance questions as autonomous AI systems begin to access, process and share sensitive information. It also opens real opportunities for organizations that can demonstrate control across organizational and technological boundaries, along with the responsibilities that come with them.
Key Takeaways
- Data sovereignty covers who can access, exchange and use information, and under which law. Storage location answers only part of that question.
- The 2026 Verizon Data Breach Investigations Report found a third party involved in 48% of breaches, which makes control after information leaves the organization a core security requirement.
- AI agents and unsanctioned AI tools widen the circle of systems that touch confidential data, so access boundaries, processing location and key control need explicit governance.
- Switzerland's 2026 sovereignty agenda, from the Swiss Government Cloud tender to new rules on cross-border data requests, shows policy moving from hosting location toward control.
Sovereignty is a question of control, beyond location
For most of the past decade, the sovereignty conversation centered on data residency: which country hosts the servers and which region the cloud contract names. Residency still matters, particularly for organizations subject to Swiss or EU data protection law, yet it describes only where information sits at rest. Sovereignty asks a broader set of questions about which laws apply to the data, who operates the infrastructure, where administrators are based, who holds the encryption keys and which third parties can reach it, as Mani Nagasundaram of MariaDB argued in Interface in late October.
Our earlier analysis of what digital sovereignty is and why it matters more in 2026 describes the same shift across three dimensions, legal, technical and operational, and traces how the question has moved from where data is stored to who can actually read it, under whose law and at whose request. The US CLOUD Act remains the most cited reason the legal dimension matters, since it allows US authorities to compel US-based providers to disclose data under their control regardless of where that data is stored.
For security and compliance leaders, the practical consequence is a change in the questions they put to vendors and to their own architecture. Asking where data is hosted produces a reassuring answer quickly. Asking who can read it, and on whose authority, produces a more useful one and often reveals dependencies that a hosting map conceals. Those dependencies multiply the moment information leaves the organization.
The exchange problem begins at the organization's edge
Most organizations run reasonable controls for confidential information inside their own perimeter, with identity management, role-based permissions and monitoring. The picture changes when a due diligence file goes to an external adviser, a claims document to a loss adjuster or a tax statement to a client. The 2026 Verizon Data Breach Investigations Report, which covers incidents from November 2024 to October 2025, found a third party involved in 48% of breaches, up from 30% the year before, a rise of 60%. Third-party data risk has moved from a procurement checklist item to one of the main routes by which confidential information is exposed.
The weak point is usually the channel. Email attachments create copies the sender can no longer recall, and generic file-sharing links often outlive the business relationship they were created for. Once a document sits in a partner's inbox, its owner has little visibility into who opened, forwarded or stored it, while regulatory accountability stays with the organization that shared it.
Secure file sharing with external parties therefore depends on properties that remain in force after a document leaves: access tied to a verified identity, the ability to expire or revoke that access, an activity log that records what happened, and encryption that protects the content along the way. The same discipline applies to customer communication, where a secure customer mailbox or digital document delivery channel replaces unprotected email for statements and contracts. It also applies when relationships end, since closing access cleanly when employees leave is part of the same control model.
AI agents widen the circle of systems that touch confidential data
Agentic AI adds a different kind of boundary crossing. An assistant that summarizes a contract or an agent that drafts a client response needs access to the underlying documents, and every connection it uses becomes a path along which confidential information can travel. The Verizon report found that shadow AI, meaning employees using unauthorized generative AI services, has become the third most common non-malicious insider action in its data loss prevention dataset, a fourfold increase year over year.
Sanctioned tools raise their own questions. Since April 17, 2026, Microsoft 365 Copilot has by default allowed language model processing for EU and EFTA tenants to move outside the EU Data Boundary during peak demand unless an administrator switches the setting off, as our CISO Antonio Mecci explains in When your AI leaves the country without telling you. Storage stays where the contract says, while processing follows available capacity.
Governing AI and confidential data starts with principles that data protection law has long required, including purpose limitation, data minimization and a clear legal basis for processing. For agents, those principles translate into three operational questions:
- which information the agent can reach and under whose identity,
- where that information is processed when the model reads it, and
- who holds the keys that make the content readable at all.
The third question carries the most weight. When encryption keys stay with the data owner, a request to the service provider, whether it comes from an attacker, a misconfigured integration or a foreign authority, yields encrypted content, and an agent can work only with what a key holder has deliberately made available. SecureSafe is built on a zero-knowledge architecture for this reason.
Switzerland is turning sovereignty into policy
The Swiss debate shows how quickly the topic is moving from principle to procurement. In December 2025, the Federal Council made digital sovereignty one of three focus topics of the Digital Switzerland Strategy 2026, alongside Switzerland's role as a digital host state and the introduction of the e-ID. Its report on digital sovereignty defines the term as the capacity for control and action in the digital space that the state needs to fulfill its tasks.
Implementation followed during 2026. In September, the Federal Chancellery launched a program for sovereign workplace software to test whether open-source tools can run core office functions for the federal administration, and in October it designated the program a key project. On October 20, the Federal Office of Information Technology, Systems and Telecommunication opened the Swiss Government Cloud tender, with a maximum volume of CHF 7 billion over 14 years, according to Swiss IT Magazine.
A second development illustrates why hosting location settles only part of the question. On October 28, the Federal Council adopted an interim solution under which Swiss providers can obtain a special permit to hand electronic evidence directly to EU law enforcement authorities, including data stored in Switzerland, to resolve a conflict between the EU's e-evidence rules and Swiss criminal law, as Swiss IT Magazine reported. The measure is formal and time-limited, and it underlines that Swiss data hosting answers where information sits, while access rules and key control decide who can read it.
Conclusion
Sovereignty, secure data exchange and AI governance are three views of the same requirement: confidential information has to remain under its owner's control wherever it travels and whichever system processes it. That requirement is becoming a procurement criterion and an audit question, and regulated buyers increasingly expect vendors to demonstrate how access, exchange and key control work in practice.
Organizations that map how their most sensitive information actually moves, to partners, to customers and into AI tools, will be better placed to answer those questions before an auditor or regulator asks them.
Enabling the secure and controlled use of information across organizational and technological boundaries is where the next phase of cybersecurity work begins, and it is where the industry's opportunities and responsibilities meet. Helping organizations keep that control as confidential information moves between their teams, partners and customers is the focus of our work at SecureSafe.
Related Articles
Why businesses still need a password manager in 2026 and beyond
Attackers have found new ways in, but stolen credentials still turn a small intrusion into a serious breach. This guide explains why password managers remain a core control for teams and how they help end access cleanly when employees leave, without losing the credentials those employees shared with colleagues.
Phishing in the C-Suite: How Executives Become the Security Gap
AI-written email, cloned voices, and deepfake video calls are raising the stakes of executive phishing. This article looks at what the 2026 data shows about the threat, what a successful attack costs, and the measures leadership teams can put in place to reduce their exposure.