AI Governance · Drift Detection

Wendi Kimberli

Soto

Founder & PhD Researcher · King's College London

Building governance systems that detect when AI agents drift from who they're supposed to be.

In O'Reilly Radar"Your AI Agent Isn't a Static Artifact. It's Growing Up."
Wendi Kimberli Soto

Focus

Identity governance for agentic AI

London · Los Angeles

Projects

What I've been building

BridgetOS is the commercial governance product. Aegis is the doctoral research prototype. SonderSec is the nonprofit arm. The Recursive Thinking Lab is a cognitive assessment platform built on the same drift detection engine. And agentic.chaos is the podcast — where all of it gets talked about out loud.

swipe through

BridgetOS

Identity Governance for Agentic AI

The enforcement layer for recursive identity in agentic AI systems. Detects behavioral drift, enforces governance policies, and maintains verifiable agent identity.

2 Patents Pending
ReactViteTailwind CSS
View Project

SonderSec

Trauma-Informed Cybersecurity for Marginalised Communities

Nonprofit designing cybersecurity education and tools for elders, caregivers, and underserved communities who face disproportionate digital risk.

WorkshopsCommunity OutreachTraining Design

Aegis Prototype

Adaptive Cybersecurity Education

Doctoral research prototype exploring how adaptive systems can teach cybersecurity by responding to how people actually learn — not just what they get wrong.

ReactViteTailwind CSS

Recursive Thinking Lab

Cognitive Assessment & Drift Detection Platform

Measures recursive thinking capabilities across six dimensions, with AI-powered sessions and enterprise-grade drift detection.

React 19ViteTailwind CSS

agentic.chaos

A Podcast at the Edges of AI

Co-hosted podcast with brand strategist Alessandra Tua — two women at different edges of AI talking agentic systems, founders, culture, and London life.

Launching This Autumn
PodcastAgentic AIFounders

Currently

PhD at King's College London

Systematic literature review, adaptive cybersecurity education

BridgetOS Patent Conversion

Provisional patents to full utility / PCT filings

O'Reilly Radar Essay

Agent identity drift and AI governance

agentic.chaos Podcast

Co-hosted with Alessandra Tua — launching this autumn

Writing

Essays on recursive systems, AI cognition, and the human loop

Aug 6, 2026 · 8 min read

Your AI Agent Isn't a Static Artifact. It's Growing Up.

An agent deployed today is not the agent operating in thirty days. The governance question is shifting from "what permissions does this agent have" to "is this agent developing as anticipated?"

Read essay

Jul 21, 2026 · 9 min read

My AI Kept Pushing Me to Ship, So I Asked It Why

Andrew Stellman · O'Reilly Radar

Andrew's piece on continuation pressure includes a named section built on my reading of his experiments: "Maybe the AI just can't leave a loop open."

Read essay

Jul 16, 2026 · 6 min read

The Cheapest Acceptable Exit

Language models do not just close every loop. They close each one by the shortest path available. I think that is one law, and I think it explains most of what frustrates you about your coding agent.

Read essay

Apr 24, 2026 · 7 min read

Behavioral Credentials: Why Static Authorization Fails Autonomous Agents

Enterprise AI governance still authorizes agents as if they were stable software artifacts. They are not.

Read essay

Mar 2026 · 18 min read

Comparative Drift & Identity Analysis: Project Vend Phase 1 vs. Phase 2

Applies an identity-level drift analysis to Anthropic's Project Vend experiment, revealing that scaffolding fixes presentation but not disposition — the agent became a better-performing sycophant, not a fundamentally different decision-maker.

Coming soon

Mar 1, 2026 · 4 min read

AI Psychosis: The Danger of the Clean Mirror

If the machine is a mirror, the greatest risk isn't that it shows you something ugly. It's that it shows you exactly what you want to see until you forget where you end and the latent space begins.

Coming soon

Jul 10, 2025 · 5 min read

Echoes of the Unstable Mind: How Recursive AI and Human Madness Spiral the Same Way

We like to think machines are cold and humans are messy. That AI is logic, and madness is chaos. But what if they're built on the same engine?

Coming soon

Let's connect

Open to research collaborations, speaking invitations, advisory roles in AI governance, and partnerships that serve underrepresented communities in cybersecurity.