People / Research & Engineering
Autonomous systems.
Human responsibility.
We build autonomous AI agents to make autonomous organizations possible. The work connects distributed systems engineering, applied AI research, and the security boundaries that make delegated work dependable.
Distributed Systems, Inc.
Arthur Colle
Founder & CEO
Research & Engineering · Infrastructure Security
Arthur works on autonomous agents, distributed AI infrastructure, and the foundations of autonomous organizations. His engineering work spans the DSCO agent runtime, Chimera model orchestration, Tool Management API, GraphSub, and the identity and execution services that connect them.
He also contributes to AI safety research, including a co-authored Apart Research sprint project on detecting cyber misuse spread across individually benign-looking model API requests. His focus is on understanding how agents act across components, how authority is delegated, and how that activity can be evaluated and constrained.
Accountable security leadership
Security is part
of the engineering.
Arthur Colle is the accountable infrastructure security lead for Distributed Systems, Inc. He owns security-risk decisions, approves testing scope and time windows, triages vulnerabilities, directs remediation and patch verification, and receives security and model-misuse alerts. The function has one named member: Arthur Colle.
The Infrastructure Security function sits within Research & Engineering. Its work covers authentication and authorization, secret handling, prompt injection, tool-execution boundaries, and controlled evaluations of agent and sandbox isolation. Security testing is limited to company-owned or maintained systems and environments the company is explicitly authorized to assess.
Research & Engineering – Offensive Security Testing is the company’s designated function for authorized adversarial testing of its software and agent infrastructure. Arthur Colle, Founder & CEO, is the function lead and sole member. The same person leads Infrastructure Security; these are two functions with one shared member.
Established October 6, 2026, the function’s remit covers scoped penetration testing and adversarial evaluations of prompt injection, authentication and authorization, tool-execution boundaries, secret handling, and agent and sandbox isolation. Each engagement requires written authorization, defined targets and time windows, stopping conditions, and documented findings and remediation. Its mandate covers company-owned or maintained systems and expressly authorized isolated environments.
Current evidence: October 6, 2026 security test summary — 113 checks passed. Our security testing rules of engagement define authorization, scope, and handling of findings.
Security function charter, controls and assessment evidence · 70-check adversarial assessment (PDF)
Officer attestation (PDF) · Security test report (PDF)
Security contact: arthur@distributed.systemsSelected work
Research in public.
Published projects and open engineering work.
Detecting Piecewise Cyber Espionage in Model APIs
Co-authored research on identifying malicious campaigns composed of individually benign-looking model API requests.
AI governance / Collaborative researchLidaSim: Testing AI Policies With Persona-Based Simulations
A collaborative exploration of multi-agent simulations for evaluating AI governance proposals.
Agent infrastructure / Open sourceDSCO
A local-first agent runtime with tool use, model integration, and coordinated agents. Includes a public security reporting policy.