DWG 01 — POSITION

Daniel A. Bauer

Security leader for the AI era.

Security architecture at carrier scale, and a seventeen-agent security program built and governed inside it.

FIG 01 — THE FLEET, AS GOVERNED

flowchart LR
    TI["<b>Threat intelligence</b>"]
    VP["<b>Vulnerability posture</b>"]
    RW["<b>Regulatory watch</b>"]
    FLEET["<b>17 agents</b><br/>NIST CSF 2.0 · weekly cadence"]
    DRAFT["<b>Draft only</b><br/>read-only and propose-only<br/>credentials"]
    GATE{{"<b>Human gate</b><br/>named accountable<br/>60-second kill switch"}}
    COMMIT(["<b>Commit</b><br/>people only"])
    TI --> FLEET
    VP --> FLEET
    RW --> FLEET
    FLEET -. "proposes" .-> DRAFT
    DRAFT --> GATE
    GATE == "approved" ==> COMMIT
    classDef core fill:#EAF2FB,stroke:#1667b8,stroke-width:1.5px,color:#16324f
    classDef rail fill:#FFFFFF,stroke:#1667b8,stroke-width:1.2px,stroke-dasharray:5 4,color:#16324f
    classDef gate fill:#0e7a4e,stroke:#0e7a4e,stroke-width:1.5px,color:#FFFFFF
    classDef store fill:#FFFFFF,stroke:#16324f,stroke-width:1.4px,color:#16324f
    class TI,VP,RW,FLEET core
    class DRAFT rail
    class GATE gate
    class COMMIT store

DWG 02 — IN ONE PARAGRAPH

Twenty-seven years on one network, through Alltel and into Verizon: senior network security engineer to Associate Director of Security Architecture, leading the team that plans network security services for one of the largest networks in the world. Eleven engineers, eight direct reports, a servant-leadership model, and a Credo Award. My team and I built the seventeen-agent fleet above, and I wrote the framework that governs it, so the organization moved from talking about AI-assisted security to running it on a weekly cadence, with people holding every decision that matters.

DWG 03 — SELECTED WORK

03.1

Agentic security, running weekly

Seventeen agents, NIST CSF 2.0: threat intelligence, vulnerability posture, regulatory watch.

03.2

Governance that makes it safe

ISO/IEC 42001 framework: models register before use, every agent has a named accountable human.

03.3

Draft, never commit

MCP servers on read-only and propose-only credentials; a human gate on every state change.

03.4

Evaluations as controls

A held-out incident set seeded with metric-gaming and scope-breach traps; drift monitoring.

03.5

Post-quantum at scale

100,000+ assets against NIST IR 8547 and CNSA 2.0; 5G key generation and HSM storage.

03.6

Resiliency for the 5G core

Ransomware defense: cell-site restoration exercises, first resiliency commitments for the core.

FIG 02 — POST-QUANTUM RUNWAY · NIST IR 8547

DWG 04 — PROOF YOU CAN OPEN

danbauer.net/governed-agents

A reference design for running an agent fleet in a security estate, beside the version I run every day.

danbauer.net/two-handed-estate

Several model families under one set of rules, and the routing between them.

github.com/chorusse/agentic-estate

A runnable agentic estate: five planes, one map, an eleven-lesson field guide. Public, MIT.

github.com/chorusse/decision-panels

A Claude skill that convenes advisors and says “the panel added nothing” when that is true.

chorusse.com/labs

The studio where the work is published, including a head-to-head that credits the rival where it is better.

DWG 05 — OPERATING PRINCIPLES

Agents draft; people commit. Every model is breachable; the architecture assumes it.

Prove it, don't claim it. Every control ships with the measurement that could show it isn't working.

A record is not the thing. Claims settle by restoring and fetching, not reading the log.

Learn before the credential exists. Self-taught ahead of the field once; doing it again with agents.