If you sell AI into government or regulated industry, the hardest question is rarely about model quality. It is about who can legally hold the data, who can be audited, and which jurisdiction’s rules apply when the two disagree. On September 16, 2026, Cohere announced it had signed a definitive business combination agreement with Aleph Alpha, following the planned partnership the two companies released in April of that year.
What the agreement actually says
According to Cohere’s announcement, the combined company will operate globally as Cohere and is described as the first transatlantic sovereign AI solution. The transaction is still subject to final regulatory approvals, and the announcement says it is expected to close later this year, with integration and product details to follow.
The structural commitments are the part worth reading closely. The company plans to be dual-headquartered in Berlin and Toronto, keep Aleph Alpha’s Heidelberg office as a research center, and grow headcount past 1,000 across both continents. Cohere states the combined entity will operate within the applicable legal and regulatory frameworks of both jurisdictions, including requirements governing sensitive data and technology, with safeguards and oversight mechanisms built into its structure.
Two leadership appointments take effect on close: Ilhan Scheer, currently Co-CEO of Aleph Alpha, becomes Chief Operating Officer, and Samuel Weinbach, Aleph Alpha co-founder and Co-Chief Research Officer, becomes Chief Research Officer. Both are contingent on completion of the transaction.
Why this is a procurement story, not a model story
Aidan Gomez, Co-founder and CEO of Cohere, framed the rationale around choice: no government or enterprise should have to pick between capable AI and control over their technology. Scheer’s quote points at the same operational reality from the other side, describing a sharpened focus on specialized language models for governments and regulated industries.
That framing matters because sovereign AI buyers are not shopping for benchmarks. They are shopping for a vendor whose legal home, data handling, and oversight chain survive an audit in their own country. A vendor anchored in two jurisdictions is a different risk profile than one anchored in a single one, and it is a different contracting conversation.
The announcement also names the infrastructure layer: Cohere says it will advance its partnership with the companies of Schwarz Group to deliver sovereign AI on STACKIT, Schwarz Digits’ sovereign cloud service. Christian Müller, CEO of Schwarz Digits, described the alliance as a way to strengthen independent AI infrastructure without forcing organizations to choose between performance and digital independence.
What builders should watch before close
The supplied announcement does not specify pricing, model availability, API changes, or which products will be offered in which market. Those details are explicitly deferred. If you are mid-procurement or mid-integration, that gap is the practical constraint: you cannot plan a migration path around a deal that has not closed.
Three things are worth tracking. First, whether the dual-jurisdiction structure changes data residency options for existing Cohere or Aleph Alpha customers. Second, whether the STACKIT partnership produces a deployment target you can actually provision against. Third, whether the leadership appointments land as announced, since both are contingent on close.
For teams building on regulated data, the broader lesson is that vendor jurisdiction is now a design input, not a legal afterthought. The same discipline applies when you are deciding how much of your stack depends on a single provider’s terms — a concern that shows up in everything from how you scope agent access to production systems to how you document data flows for review.
A grounded next step
If this deal touches your roadmap, do not rewrite your architecture yet. Instead, write down which of your current deployments would need a second jurisdiction to be viable, and ask your vendor contact directly what changes at close. The announcement is a commitment to a structure, not a shipping date for features.
Sources
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.