Engineering, one of the leading players in digital transformation in Italy with a significant international presence, has developed the IS-IA – Italy’s Sovereign Intelligence Architecture, an architectural approach based on the concept of Italian, governable and secure Artificial Intelligence.

Thanks to this innovative strategy, the company will help organizations move beyond relying solely on AI systems that are difficult to govern, inspect, and certify. Artificial Intelligence thus becomes an asset that generates cumulative value over time and contributes to building a defensible and long‑lasting competitive advantage.

IS-IA is built on EngGPT  2, an Italian platform for Private Generative AI that is controlled, efficient, secure, transparent and highly customizable to organizational needs, fully aligned with AI Act standards, and capable of delivering performance comparable to the best international open models in its class.

Aldo Bisio, CEO of Engineering, stated: “The widespread adoption of GenAI is an urgent step for the country’s competitiveness. However, it must be fully governable in order to avoid the ‘expropriation’ of distinctive skills and know‑how from our companies and institutions. We have developed a sovereign architecture, meaning  governable and inspectable, that  enables to retain full control over data, models and infrastructure, preserving governance of cognitive assets and solutions of companies  and Institutions. With IS-IA, our Italy’s Sovereign Intelligence Architecture  built on  EngGPT  2 foundation model, we bring to market a GenAI architecture developed in Italy that is efficient, open and compliant with the AI Act, capable of integrating into hybrid environments and adapting to the most regulated environments. It is not just about adopting AI: it is about building and governing it as a strategic asset for the country. IS-IA and EngGPT  2 are a concrete enabler of this vision and an opportunity to strengthen technological autonomy, innovation and growth for the country.”

Artificial intelligence is growing faster than any other technology — according to Gartner, global AI spending will reach approximately USD 3.3 trillion by 2027 — but this race contains a paradox: most AI applications today are interfaces built on the same few models, accessible to everyone and therefore unable to generate a real competitive advantage.

To overcome these limitations and enable organizations to adopt secure, efficient and transparent AI designed to adapt to their data, processes and decision‑making, Engineering has developed EngGPT  2.

IS-IA is the end-to-end architecture that Engineering makes available to companies and Public Administrations to govern the entire AI lifecycle: from infrastructure to data, from the foundation model to specialization, through to orchestration within business processes and governance. EngGPT 2 is the sovereign foundation model at its core, but IS-IA is what enables it to be transformed into cumulative value over time, through a growing library of reusable and certified specialized models. It is the difference between “using AI” and “owning your own intelligence.”

Key features of EngGPT  2

EngGPT  2 is a sovereign AI platform.

It is an Italian LLM with 16 billion parameters, trained from scratch on 2.5 trillion tokens (word/part of a word): equivalent to studying approximately 30 million books. Of these trillions of tokens, around 25% consist of “data” in the Italian language, enhancing the model’s ability to capture linguistic nuances and the economic, social, industrial and regulatory contexts specific to Italy.

EngGPT  2 also leverages the work of 130 Italian researchers from the Group’s R&I Competence Center, who in 2025 alone were involved in over 50 projects on AI and Data.

Among EngGPT  2’s distinguishing features are efficiency and sustainability.

Thanks to the advanced Mixture‑of‑Experts architecture, the model activates only the components required for each request. As a result, EngGPT  2 can use up to one‑fifth of the energy required by equivalent models, leading to a lower environmental impact, also due to reduced water consumption for data‑center cooling.

In terms of energy efficiency, it is also worth noting that EngGPT  2 customers will access the platform also via Engineering’s three data centers (located in Pont‑Saint‑Martin, Turin and Vicenza), all powered entirely by renewable energy. The Pont‑Saint‑Martin facility in the Aosta Valley is equipped with a geothermal cooling system that also uses groundwater to naturally cool systems, through a closed‑loop circuit that avoids waste and preserves the area’s hydrological balance.

EngGPT  2’s efficiency also translates into:

• Training up to 10 times more efficient than dense models of comparable size (models that activate all parameters for every user request)
• A 50–80% reduction in inference costs compared to dense models of comparable size, i.e. in generating answers from new inputs.

Designed according to rigorous security‑by‑design principles, EngGPT  2 ensures full traceability and auditability.

The platform is verifiable and compliant with the AI Act, as it allows visibility into which data is used, how it was trained, with which techniques and with which limitations.

These characteristics also make the platform suitable for highly regulated sectors such as finance, healthcare, energy and Public Administration.

EngGPT  2 delivers open AI that hides nothing, verifiable, transparent and publicly available.

Being “open” means making the model understandable and verifiable: everything is disclosed — weights, data sources, training processes and results. Datasets, pipelines, techniques, metrics and their limitations are available in a dedicated technical report, also publicly released, ensuring maximum transparency and full verifiability of the training process.

EngGPT  2 features an architecture designed for easy interoperability with other platforms.

This enables agentic solutions that combine proprietary models with market ones, while protecting intellectual property (IP), enhancing distinctive know‑how and adopting specialized, customizable AI solutions.

In this context, EngGPT  2 can integrate with:

• Agent platforms via APIs, enabling orchestration within multi‑agent automation processes and native integration with AgentStudio, Engineering’s proprietary agentic platform
• Data analytics platforms, ERP and CRM systems, enabling data analysis, insight extraction and real‑time recommendations scenarios applicable to various industry-specific processes.

EngGPT  2 is also available as a service on major cloud marketplaces, allowing enterprises to integrate the solution into their existing infrastructure.

IS‑IA – Italy’s Sovereign Intelligence Architecture is Engineering’s approach to sovereign, secure and sustainable AI. Built on EngGPT  2, the Group’s Italian foundation model, IS‑IA enables enterprises and Public Administration to govern data, models and infrastructure, transforming Artificial Intelligence from a commodity into a strategic asset. IS‑IA is AI — but it is Italian.