Shadow AI
Why data sovereignty is becoming a prerequisite for AI
In the summer of 2026, Switzerland finds itself in a phase of hyper-adoption amid a governance vacuum. While 77 % of the Swiss workforce already uses AI regularly in their daily work, more than half of them disregard applicable company policies and upload sensitive data to unsecured LLMs.
This so-called shadow AI is shaping up to be the biggest cybersecurity challenge in the coming years: In May of this year, every 25th AI prompt already posed a high risk of business-critical information leakage.
The good news is: We have the technological answer to this loss of control.
Centraya AI offers the secure path to scalable AI adoption by acting as an upstream security and encryption layer.
Instead of hindering innovation out of uncertainty, Centraya AI enables the use of cutting-edge cloud LLMs without confidential content like PII (Personally Identifiable Information) leaving your own control zone unprotected.
The core of our solution is prompt encryption. Sensitive information in prompts is encrypted before it is transmitted to external models. This allows AI-powered workflows, from the analysis of patient data to the creation of confidential documents, to be implemented securely and with control, even in highly regulated sectors such as medicine or law, and in compliance with the Swiss Federal Act on Data Protection (FADP) or the GDPR.
Why data sovereignty is becoming a prerequisite for AI
Our dialogue format is coming to Switzerland.
Exchange, new products, and best practices in Zurich and Geneva.
Why new releases need to become usable faster and more controlled.
Why data sovereignty is becoming a prerequisite for AI
Our dialogue format is coming to Switzerland.
Exchange, new products, and best practices in Zurich and Geneva.
Why new releases need to become usable faster and more controlled.
AI has arrived in companies and become an everyday tool before governance, security, and compliance could react to it.
While employees may seem to increase their efficiency in information processing, text creation, or process flows, shadow AI can quickly become a risk for organizations.
Uncontrolled outflow of sensitive data is the key phrase.
Michael Rieder, Head of IT Cloud Consulting at e3, provides context in an article in Netzwoche's special cybersecurity edition and explains why traditional security mechanisms are reaching their limits when dealing with AI. His thesis: data sovereignty is a prerequisite for secure innovation.
For companies, it is essential to know the conditions under which they can use artificial intelligence without losing control over their business-critical information. Bans alone are insufficient. Architectures that anchor information protection directly to the data are in demand: with DLP, granular controls, encryption, and security mechanisms that engage before sensitive content leaves the company.
After successful events in Frankfurt and Munich, the Peer Dialogue for Cyber Resilience & Security e3 Perspectives is now coming to Switzerland for the first time.
The event series is deliberately designed to be dialogue-oriented. In small groups, experts, executives, and subject matter experts discuss real cyber incidents, decision dilemmas, and proven solutions at eye level. The goal: to take away concrete starting points for greater data sovereignty, traceability, and resilience.
We look forward to welcoming you to one of these events and discussing with you. Save the date in your calendar today!
» Learn more about e3 Perspectives
Once again in 2026, we, together with Symantec and Arrow, invite you to our User Groups in Zurich and Geneva. The focus will be on current developments, best practices, and practical experiences related to DLP and network security solutions in everyday business.
We are now also opening the afternoon to interested professionals seeking to network with e3, Symantec, and peers.
This dialogue has proven to be particularly valuable in recent years: in addition to a direct insight into new products, our customers appreciate the open exchange among themselves.
e3 DLP receives a new user interface. The revised interface will simplify incident management and will include, for example, lists and panels in the Case Detail Views. This will support security teams in identifying critical security incidents more effectively and quickly, and taking appropriate action.
The first innovations of the new UI are included in preview mode starting with release 12.1.2. Starting with version 12.2.0, planned for mid-August, lists can be personalized. Personalization of the panels in the Case Detail Views is scheduled for version 12.2.3, which is expected to follow later in the year.
LLMs like ChatGPT would often be particularly helpful in everyday work if they could work with real context. However, this context is often confidential because customer, patient, client, or company data must not be shared unprotected with external AI systems.
With Centraya AI, we've created an upstream security and encryption layer for using open cloud LLMs. Sensitive content is encrypted before it reaches the AI. This allows teams to use AI productively without exposing confidential information in clear text.
Centraya AI is now available for pilot projects and live demos.
A recent post by Broadcom shows that AI accelerates the identification and analysis of vulnerabilities, both for companies and their security teams, and for potential attackers. This makes the speed of response with which security updates are evaluated, released, and implemented in production environments a crucial security factor.
From our work with clients, we know that the adoption of new software versions into production often takes a lot of time due to complex processes, technical dependencies, and necessary checks. At the same time, the period between the discovery of a vulnerability and its potential exploitation is shortening. Efficient and reliable release processes are therefore becoming a central component of effective cybersecurity.
Our goal is clear: we want to make new manufacturer releases available to our customers faster and more reliably. To achieve this, we continuously optimize our testing and release processes, evaluate new versions more efficiently, and specifically secure existing functions. This enables faster deployment without compromising on quality, stability, or necessary analyses.
DE DE 
EN
FR
