Engineering takes sovereign AI to Brussels: data control beyond model size
In the proposal Engineering brought to the European Parliament, AI sovereignty means the ability to choose and govern systems rather than technological isolation.[1]
The company’s 10 November release and an Adnkronos report published by Il Tirreno at 14:00 document the IS-IA discussion, co-hosted by Parliament Vice-Presidents Pina Picierno and Antonella Sberna.[1][2]
This was a company-promoted discussion, not the adoption of a new European programme.[1]
## What the architecture offers
IS-IA stands for Italy’s Sovereign Intelligence Architecture: Engineering describes an approach developed on its proprietary private GenAI platform EngGPT 2, intended to keep data, models and knowledge within controllable environments.[1]
CEO Aldo Bisio presents modularity and the orchestration of different models as ways to match systems to specific organisational needs.[1]
Bisio argues that post-training specialisation can enable more agile models to deliver comparable results using fewer resources.[1][2]
That is the executive’s position: the sources consulted provide no independent benchmark demonstrating this across applications. The agency report records what was said; it does not certify the product’s performance, security or cost savings.
## Security did not launch today
The discussion also included EngSure, the group’s agentic cybersecurity platform.[1][2]
This was not a 10 November launch: Engineering’s original release is dated 15 October 2026 and already placed EngSure within its IS-IA strategy.[3]
According to Engineering, its four modules cover monitoring and response, attack simulation, converting technical evidence into information for governance and audit, and analysing the provenance and reliability of data and models.[3]
The company also describes controls that can block agents acting beyond their authorisation or require human intervention; these are supplier claims, not features verified here in an operational test.[3]
## Questions before procurement
Our editorial view is that “sovereign” should not be treated as a self-sufficient certificate. Buyers should ask where data is processed, who can access it, how models can be replaced and which actions require human approval. These are suggested evaluation questions, not requirements adopted at the meeting.
The verified development is the Brussels discussion on 10 November, not the first invention of IS-IA or an endorsement by the European Union.[1][3]
Keeping these levels separate allows an industrial proposal to be reported without turning it into an institutional guarantee.
## Sources
[1] https://www.eng.it/en/news/press-releases/2026/11/engineering-porta-a-bruxelles-la-sfida-europea-per-un-ai-sovrana — Engineering, evento Bruxelles 10 novembre 2026
[2] https://www.iltirreno.it/ultimora-adnkronos/2026/11/10/news/ia-engineering-porta-a-bruxelles-la-sfida-europea-per-una-sovranita-tecnologica-1.100940883 — Adnkronos via Il Tirreno, 10 novembre ore 14
[3] https://www.eng.it/en/news/press-releases/2026/10/engineering-lancia-engsure--la-piattaforma-agentica-per-la-cybersicurezza-AI — Engineering, lancio EngSure 15 ottobre 2026