Centraya AI: Safe use of AI for SMEs

The experimental phase in dealing with artificial intelligence is over. In 2026, success will no longer be determined by the number of AI tools tested, but by the ability to integrate artificial intelligence into existing business processes in a controlled, compliant and productive manner. Anyone who treats AI merely as a technical toy or slows it down completely out of uncertainty risks being at a competitive disadvantage.

Security is not an obstacle, but a prerequisite for scaling AI and the associated productivity gains in the company in the first place. Only an architecture that considers security from the outset will enable the full potential of global language models to be exploited with confidence.

Increasing compliance requirements

Even without a dedicated AI law, the use of AI in Switzerland is already subject to existing requirements - particularly in terms of data protection law, transparency obligations and data protection impact assessments for risky applications. In highly regulated sectors such as finance, law or healthcare, additional specific requirements also apply.

The crucial question is therefore not whether AI should be used, but under what conditions it can be used safely, responsibly and with real business added value.

Protection of sensitive data in the prompt

This is exactly where Centraya AI comes in. The solution enables companies to use modern open cloud language models productively without releasing sensitive information unprotected from their own control zone. As an upstream security and encryption layer, Centraya AI already protects confidential content in the prompt, creating the basis for secure, controlled and compliant AI use in regulated and sensitive environments.

Policy Factory for precise data recognition

Alongside the launch of Centraya AI, E3 is introducing the Policy Factory. This new service bundles all relevant services for the development, revision and optimization of DLP (Data Loss Prevention) policies. The quality of a DLP solution is already determined by the question of which information is classified as worthy of protection and how it can be reliably identified.

In practice, it has become apparent that existing policy collections have grown over the years, are often no longer adequately maintained or are not precisely tailored to today's requirements. The development of truly resilient search patterns requires specialist knowledge, a methodical approach and an understanding of how regulatory, organizational and technical requirements interact.

Data sovereignty as a strategic necessity

In 2026, Switzerland is no longer just an observer of global tensions, but increasingly part of an insecure and volatile environment. Espionage, disinformation and cyberattacks affect companies and institutions more directly and more frequently today than just a few years ago. At the same time, familiar securities are coming under pressure. In a world of strategic uncertainty, international agreements and contractual regulations no longer offer the same protection that companies were previously able to rely on.

The question of data sovereignty and access control is particularly relevant here. Extraterritorial regulations such as the US Cloud Act can have far-reaching consequences depending on the constellation - even if data is not stored in the US. The risk from within is also growing: infiltration by shadow AI has long been a reality in Swiss companies. Employees use AI chatbots so naturally that conventional security measures are being undermined and large amounts of data are flowing to global hyperscalers.

Where is sensitive information stored, who can access it and how are authorizations and key concepts organized? Today, data security goes far beyond technical issues and has become a central strategic management task.

More information: www.e3.ch