Kristen Smith 
- Cloudera research found that only 16 percent of public sector respondents can access all their data when needed
- Brooks breaks AI governance into models, data and agents
- Cloudera Government Solutions holds FedRAMP Moderate authorization
Agencies that wait until they are ready to scale before deciding how to govern artificial intelligence models, data and agents will hit roadblocks, according to Cloudera‘s Ian Brooks.
The director of global industry AI solutions for government led a session on AI-assisted software development at the 2026 National Laboratories Information Technology, or NLIT, Summit. He revisited the event in a Q&A published on the company’s website Thursday.
Brooks argued that security requirements in classified environments will not relax as AI advances, and that agencies handling sensitive workloads need models that operate where the data already sits rather than moving data to an external model.
Agencies continue to weigh how data infrastructure and governance shape AI adoption. The Potomac Officers Club will host the 2026 FedCiv Summit on Oct. 29 at the Ritz-Carlton Pentagon City in Arlington, Virginia, where leaders from federal civilian agencies and industry will discuss AI adoption, cloud and data infrastructure, workforce strategies, and cross-agency initiatives. Save your seat now.
What Does Brooks Mean by Private AI?
Under the private AI approach, an organization hosts its own large language models on infrastructure it controls, runs them through its own inference service and manages access to the model endpoint.
Brooks said the approach also lets agencies use open-source models suited to a particular task without moving sensitive information outside the environment. He cited control over infrastructure and costs as another reason, in contrast to per-token pricing from outside providers.
His NLIT demonstration connected a development environment to a private inference service, showing coding assistants running without the code and prompts leaving organizational control.
Why Does Brooks Emphasize Data Readiness?
Data foundations are where Brooks sees agencies running into trouble as AI adoption spreads. Cloudera research found that nearly 80 percent of enterprises say limited data access holds back their AI and data projects, while only 16 percent of public sector respondents say they can access all their data when needed.
Solving this issue demands an end-to-end focus on the data lifecycle — from ingestion and scalable management to security, governance and AI readiness.
Scope of Governance
According to Brooks, federal AI governance must encompass three critical layers:
- AI models: Tracking deployment environments, operational runtime and user access permissions.
- Underlying data: Restricting model visibility to align strictly with role-based security requirements.
- AI agents: Managing autonomous systems that execute tasks within third-party environments, introducing novel security risks beyond standard generative text models.
What Federal Authorizations Does Cloudera Hold?
Cloudera earned Federal Risk and Authorization Management Program Moderate provisional authority to operate in June 2025, enabling agencies to handle Impact Level 2 federal information on the Cloudera for Government platform via AWS GovCloud. The company previously secured agency authorization at the moderate level in December 2024 through the Department of Veterans Affairs’ sponsorship.
