IBM is adding its own twist to Sovereign AI with a technology preview next month of a product that acts as a cloud-agnostic control plane for deploying AI applications.
Sovereign AI typically refers to generative AI services that operate within specific geographic or legal boundaries in accordance with local regulations. Over the past year, interest in sovereign AI has increased as political instability and regulatory divergence have surfaced.
But previous sovereign AI efforts have primarily focused on storing data used by AI in specific regions, rather than controlling where the rest of the systems involved in AI applications reside, Will Streit, IBM’s vice president of software, said in an interview with Informa TechTarget.
“The compliance controls you want to verify in production may be different than in development,” Streit says. “So how do we extend that to AI within those workloads?”
IBM Sovereign Core is planned as an answer to that question, Streit said. It has been in development with design partners over the past few months and will be available as a limited technology preview to a wide range of customers in early February. Streit says the product can be installed with one click and up and running in as little as a day. It connects to existing software development and deployment systems through a container registry, allowing users to choose which underlying system to use.
“The generative AI angle is adding a new wrinkle to sovereignty in general,” said Jason Andersen, an analyst at Moore Insights and Strategy. “Up until now, regulators have wanted to know where data is being stored and processed, but the data that’s being produced is a little different. It’s starting to get a little weird because the data is being created. [on the fly]. ”
Agents are like children, running around and causing havoc. How do you control it at runtime rather than after the fact?
jason andersenAnalyst, Moorish Strategies and Insights
IBM Sovereign Core could appeal to large enterprises subject to multiple territorial regulations that need a comprehensive software-based system to bring compliance controls and validation to their generative AI inference and application deployment processes, Andersen said.
“When you deploy a generative AI application, you want to make sure that something is sovereign before it is generated so that the results are not placed on a random GPU,” he said. “You have to start thinking about this from an agent’s perspective as well, and that’s where things get really weird.
“Agents are like children. They run around and cause havoc,” Andersen said. “How do you control it at runtime rather than after the fact? It’s not like checking in and checking out code.”
Andersen said the timing of IBM’s preview is due to the general availability of Oracle Cloud Infrastructure’s Generative AI service this week in Oracle’s Top Secret cloud region. AWS also launched new European cloud services this week.
“Oracle Top Secret Cloud is OCI-only, not some kind of all-encompassing hybrid,” Andersen said. “IBM may be trying to capture the messaging moment, because if you’re an enterprise and you’re running something here and there, you’re looking at 10 different dashboards to get answers.” [about compliance]And someone can change the rules at any time. ”
Sovereign AI Hybrid Cloud Tradeoffs
For now, details regarding the specific technical components and features of the tool have not yet been revealed. Existing IBM and Red Hat software product capabilities, such as Red Hat OpenShift, will serve as the infrastructure foundation for the sovereign core, while HashiCorp’s Vault will contribute access management functions within the sovereign region, Streit said. Elements of the IBM Concert Compliance Center perform compliance verification and certification. Sovereign Core also includes templates called accelerators that set up systems in compliance with certain regulations, such as the EU’s Digital Operational Resilience Act.
A complete list of product features and their origins in the IBM portfolio is not yet available, but licensing these contributing products is not a prerequisite for using Sovereign Core, Streit said.
IDC analyst Rick Villers said the initial version of Sovereign Core will likely be built on tools from IBM and Red Hat and evolve into a more flexible framework that can also integrate with third parties.
“I think the first version will be, ‘Here are the sizes and three colors for the T-shirt,’ but at some point you can add your own logo or define the material,” Villers said. “But part of the goal that IBM is trying to achieve is consistency and reproducibility.”
However, Villars said there are trade-offs in consistency between the software level and the cloud infrastructure level.
“Part of what cloud providers bring to the table is that they are pre-built, pre-standardized, and the operational issues are taken care of by the provider, so you can understand performance,” he said. “IBM is delivering, ‘Here’s everything you need, but you need to actually do it.'”
