# Agentic AI and AI security for public-sector Washington, DC Federal agencies, contractors and nonprofits in the capital need AI agents that are useful, explainable and inside an authorised boundary. We design and secure agents that run on government clouds, follow NIST guidance and keep a human in charge of every material decision. ## Key facts - **Served from:** Our global delivery team, with on-site delivery in Washington, DC - **Deployment options:** AWS GovCloud (US); Microsoft Azure Government; Google Cloud with Assured Workloads; on-premises stacks for classified or air-gapped work - **Regulations we design for:** FedRAMP-authorised cloud services; NIST AI RMF; NIST SP 800-53 controls; OMB guidance on federal agency use of AI; FISMA ## What we deliver - [Government services agents](https://www.nitronedge.com/industries/public-sector/): Case-handling agents for agencies, with a named official approving every outcome. - [Sovereign, in-country AI](https://www.nitronedge.com/solutions/agentic-ai-governance/sovereign-on-prem-ai/): Agents inside your authorisation boundary on GovCloud, Azure Government or air-gapped hardware. - [AI security](https://www.nitronedge.com/services/ai-security/): Controls mapped to NIST SP 800-53, plus red teaming against prompt injection. - [AI governance](https://www.nitronedge.com/services/ai-governance/): Agent inventories and risk records that follow the NIST AI RMF. ## FAQ ### Do you have an office in Washington, DC? We don't have a Washington, DC office. Delivery runs remotely from our global team, with on-site visits for key milestones and local partners where needed. Every enquiry gets a reply within 6 hours. ### Is NitronEdge FedRAMP authorised? No. We are not a cloud service provider. We deploy agents on FedRAMP-authorised clouds such as AWS GovCloud and Azure Government, inside your existing authorisation boundary. ### How do you keep humans in charge of agent decisions? Agents prepare cases, drafts and recommendations, and an accountable officer approves anything material. Every step is logged so decisions can be explained later.