Salesforce has expanded Missionforce with a policy engine and new OpenAI and NVIDIA capabilities, giving US government teams a more controlled way to configure AI agents for mission work. The September 16 announcement matters less as a model partnership than as an attempt to make permissions, approved data and auditability part of the agent runtime.
| Element | What was announced |
|---|---|
| Policy controls | A policy engine for agent permissions and approved data access |
| Model support | OpenAI models |
| Infrastructure and models | NVIDIA capabilities |
| Disclosure date | September 16, 2026 |
Salesforce describes Missionforce as its defence-focused agentic AI platform. Its expansion connects the platform to OpenAI models and NVIDIA technology while adding controls intended for sensitive government workflows. TechRadar independently reported the expansion and its focus on US government missions.
Missionforce puts policy before model choice
The central mechanism is the new policy engine. According to Salesforce, administrators can define which tools an agent may invoke, which approved information it may use and which actions require oversight. That shifts governance from a document beside the system into the system’s execution path.
In plain terms, Missionforce is being positioned as a control layer that decides what an AI agent may see and do before the model produces an answer or takes an action. OpenAI and NVIDIA expand the model and infrastructure options, but the policy layer is what determines whether an agent can fit a regulated workflow.
What OpenAI and NVIDIA add
Salesforce says OpenAI models will support Missionforce use cases, while NVIDIA contributes models and infrastructure. The announcement does not mean every agency workload will use the same stack. It gives programme owners options inside a Salesforce-led operating layer, subject to deployment and security requirements.
That distinction is important. A model can reason over a prompt, yet an operational agent also needs identity, data entitlements, tool permissions, logging and escalation rules. Missionforce is trying to bundle those surrounding controls so agencies can move from demonstrations to bounded production tasks.
The approach follows Salesforce’s broader push to make agents useful across enterprise systems. Lapaas Voice has previously examined the company’s portfolio of job-focused Agentforce agents and its connected AI stack with Google Cloud. Missionforce applies a similar orchestration thesis to government environments, where authorization and audit trails carry more weight.
The consequence for government AI buyers
Buyers should evaluate the expansion as an architecture, not a brand list. The useful questions are whether policies are enforceable at runtime, whether logs capture tool calls and data access, how human approval is inserted, and how a team changes models without weakening controls.
Salesforce has announced the components, but agencies still need to validate them against procurement, security and mission rules. No announcement can substitute for accreditation or workload-specific testing. The near-term signal is that enterprise AI vendors increasingly compete on governance and integration around models, not only on the models themselves.
FAQs
What is Missionforce?
Missionforce is Salesforce’s agentic AI platform for defence and government mission workflows, combining agents with data, integration and governance controls.
What changed in the September 16 expansion?
Salesforce announced a Missionforce policy engine plus support involving OpenAI models and NVIDIA models and infrastructure.
Does the announcement mean autonomous deployment?
No. Agencies still determine approvals, security controls and deployment scope; the policy engine is intended to help enforce those boundaries.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
