Atos Launches Genesis Intelligence, a Sovereign AI Suite That Can Run Air-Gapped, With Atos Codex for Critical Operations
Atos on October 26 announced Atos Genesis Intelligence, an end-to-end artificial intelligence suite aimed at organisations in heavily regulated and data-sensitive sectors, the Paris-based group said in a press release also covered by ActuIA. The suite brings data management, model development and deployment, orchestration, governance and security into a single ecosystem, and it can be deployed on public cloud, sovereign cloud, on-premises or in fully air-gapped environments.
Its building blocks include a data foundation with built-in quality checks, cataloguing, lineage and traceability; intelligent model routing that picks a suitable model for each request to keep costs down; an SLM Factory for developing and managing small language models that can be deployed locally; and an enterprise portal offering a marketplace of generative AI agents that can be tailored to specific business tasks. Atos says the architecture is built on open-source technologies to avoid vendor lock-in, and that the suite will be enriched through its Atos Sovereign Agentic Studios delivery model.
The centrepiece is Atos Codex, short for Center of Operations and Data EXcellence. It builds an operational ontology of an organisation, covering its entities, relationships, rules, permissions and possible actions, so that AI insights can be turned into actions carried out according to the level of human control the organisation sets. It draws on the cybersecurity and mission-critical systems expertise of Eviden, the group’s products brand, and targets critical infrastructure operators, transport providers, industry, defence agencies and public institutions.
What it means in practice
For French buyers, the pitch is control: the same tooling whether a workload runs at a hyperscaler or on a disconnected site. That puts Atos alongside other sovereignty plays in the market, such as the France-hosted managed AI from Mistral and NumSpot and the sovereign workplace from Twake.ai recommended for public bodies. Atos’s particular bet is the step after recommendations, letting governed agents act on operational systems. “Organizations are looking beyond AI experimentation and increasingly focusing on tangible business outcomes,” chairman and chief executive Philippe Salle said.
Atos did not disclose pricing or named customers. The group reports more than 51,000 employees serving over 4,500 customers in 54 countries, and the test will be whether operators of sensitive infrastructure trust an AI layer to execute, not just to advise.