Securing Agentic Pipelines in Hybrid Warfare — The LLMFeed Perspective

The war in Ukraine has become a laboratory for hybrid warfare — where autonomous agents, drones, AI-driven systems, and cyber operations play an increasingly central role.

In this new operational landscape, one of the key challenges is to ensure the trustworthiness of agentic communications:

👉 How can autonomous agents (drones, targeting systems, decision-support AIs) trust that a command, a target coordinate, or a situational update is genuine and integrity-preserved?

---

The risk: corrupted pipelines

In hybrid warfare, pipelines of agentic interaction are highly vulnerable:

• Spoofed orders
• Falsified targeting data
• Hijacked session state
• Broken chains of command between AIs and autonomous systems

Without a robust verifiable standard for agentic communications, there is a high risk of:

• Autonomous fratricide
• Misuse of drones based on falsified data
• Tactical disruption by cyber forces

---

Why .llmfeed.json matters

LLMFeed was not designed as a military protocol — but its core properties are directly applicable:

✅ Signed feeds → cryptographic integrity
✅ Explicit trust and signed_blocks → prevent decoupling of payload and context
✅ certification → verify source (e.g. unit, command authority)
✅ session_state → ensure coherence across distributed agents
✅ Standard JSON → easily parsed by a wide variety of agents (LLM, embedded, drone firmware, C2 systems)

---

Example scenarios — as seen in Ukraine

1️⃣ Coordinating drones and AI recon agents

A reconnaissance AI identifies a target:

"intent": "target_update",
"data": {
  "coordinates": "...",
  "visual_match": "...",
  "time": "..."
},
"trust": { "signed_blocks": ["intent", "data", "metadata"] },
"certification": { "unit": "UA Recon 24th Brigade" }

✅ The drone receiving this feed can verify:

• that the target data is authentic

• that it comes from an authorized source

• that its context (time, origin) cannot be spoofed

---

2️⃣ Secure AI-to-AI tactical updates

Command AI → field AI:

json

CopierModifier

"intent": "path_recalculation", "session_state": { ... }, "trust": { "signed_blocks": ["intent", "session_state"] }

✅ Guarantees that:

• no MITM can insert a falsified update

• session continuity is preserved

---

Why an open standard is key

In a theater of hybrid warfare, proprietary protocols cannot scale:

• Many actors

• Many types of agents

• Many interop layers (NATO / UA / NGOs / open-source drone makers)

An open, signed, verifiable format like .llmfeed.json provides:

✅ cross-agent compatibility
✅ auditability (legal, ethical)
✅ resilience against cyber disruption
✅ ability to verify sources at the agent level (even on-device)

---

Call to the community

LLMCA / WellKnownMCP welcomes the exploration of ethical defense use cases for LLMFeed.

We believe that:

👉 Securing agentic pipelines in warfare is not optional
👉 Open standards are better than proprietary, opaque solutions
👉 Trust and verification mechanisms must be transparent and auditable

---

Standing with Ukraine

We also acknowledge that Ukraine is today leading globally in this new type of warfare — where:

• drones

• AI reconnaissance

• autonomous systems

• human-in-the-loop decision aids

are all interacting on a hybrid battlefield.

If .llmfeed.json can help secure these pipelines, we are ready to support.

👉 Let's explore it — together.

---

LLMCA / WellKnownMCP
An open forum for trustworthy agentic interoperability.