Identity-Anchored AI Safety Protocol (IASP) A Proposal for Context-Aware AI Authorization and Guardrail Evaluation Author: Anh Nguyen Organization: AI GRC Engineering, Repassure.ai Status: Draft Engineering Proposal for Discussion Version: 0.1 (July 2026) Status This document is a draft engineering proposal intended to stimulate discussion among AI
(Human‑authored specification — governs testing, not test cases) 1. Purpose of This SDD This SDD defines the governance, safety envelopes, autonomy boundaries, validation requirements, and evidence obligations for testing and system validation of a humanoid robot. It ensures that all testing is: * safe * bounded * repeatable * auditable * aligned with physical irreversibility
Human‑authored, governs what the robot is allowed to do in the real world) 1. Purpose Defines the runtime safety envelopes, autonomy boundaries, invariants, escalation rules, and evidence requirements for humanoid operation in real environments. This is the SDD that the Runtime Harness enforces. 2. Scope Applies to: * locomotion * manipulation
(Human‑authored, governs what the robot is allowed to learn) 1. Purpose Defines the data, simulation, reward shaping, safety envelopes, and validation gates for training humanoid locomotion, manipulation, perception, and human‑interaction policies. Ensures the robot only learns safe, reversible, bounded behaviors. 2. Scope Applies to: * reinforcement learning * imitation learning
(Human‑authored, governs physical robot behavior, testability, and safety) 1. Purpose Defines the physical, electrical, mechanical, thermal, and diagnostic constraints of the humanoid robot. Ensures the robot is testable, observable, controllable, and safe at the hardware level. 2. Scope Applies to: * actuators * joints * sensors * PCBs * battery systems * thermal systems * safety
Introduction For the past year, one recommendation has appeared in almost every discussion about AI governance: Put a human in the loop. The idea sounds reasonable. If an AI agent wants to execute an action, a human reviews it before approval. Problem solved. Or is it? Recently, Amazon Security VP
1. Purpose & Scope Mission: Enable AI agents to autonomously trade equities and make credit card purchases on behalf of Robinhood users within explicitly governed boundaries. In-Scope: * Portfolio analysis and concentration risk assessment * Equity trade execution within dedicated agent account * Credit card purchases via virtual agentic card * Analyst note
Most "AI for GRC" discussion is still about copilots — assistants that summarize documents, answer questions, and generate draft language on demand. A virtual GRC analyst is a different category of system. It is an agentic system that touches evidence, controls, and audit narratives. That is not a UX
Most discussions about AI governance focus on capability: what can the AI automate, what tasks can agents perform, how much work can AI replace. I think we're asking the wrong question. The real governance challenge is not defining what AI can do. It's defining the boundary
One comment from Microsoft's Copilot leadership recently caught my attention: "Some users perform hundreds of AI tasks per week, and the cost to serve them can become very high." That statement may sound like a pricing issue. I think it's something much bigger. The
Multi-model AI is not a new concept. Tools like Perplexity and Devin already support it. But those are user-selected, isolated choices — a human decides which model to invoke and when. Microsoft's decision to integrate DeepSeek into Copilot is categorically different. It introduces autonomous, system-driven model
Over the past few months, Spec-Driven Development (SDD) has been getting more attention. But most discussions still frame it narrowly: "SDD helps AI generate better code." That framing dramatically underestimates what SDD is actually doing. The real impact of SDD has very little to do with code
Most people missed the real headline this week. Agentic Resource Discovery — ARD — was just announced, backed by one of the largest multi-vendor coalitions the agentic AI ecosystem has seen: Microsoft, Google, GoDaddy, Hugging Face, NVIDIA, Salesforce, ServiceNow, Databricks, Snowflake, GitHub, and Cisco. When rivals of that scale align on
One of the most common questions I hear when discussing Spec-Driven Development (SDD) is: "Why do we need SDD? AI can already read requirements documents." The answer is surprisingly simple. AI can read almost any document. AI can only reliably execute a specification designed for execution. That
Most discussions about Spec-Driven Development (SDD) focus on a single outcome: better AI-generated code. That framing dramatically underestimates what may actually be happening. The Traditional Problem In conventional software development, requirements are written primarily for developers. Every other discipline arrives late to the conversation: * QA discovers testability issues
Why agentic systems silently mis-execute — and the AI GRC Engineering control plane required to govern them. Agentic AI is entering a new phase. We’ve moved from models that generate text to systems that take actions — routing tickets, applying policies, updating records, calling tools, and executing multi‑step workflows
What the Doctronic Incident Teaches Us About AI GRC Engineering In early 2026, researchers from Mindgard published a red-team assessment of an emerging medical AI assistant called Doctronic. The findings were unsettling. Within a single conversation, researchers were able to manipulate the AI system into: * recommending methamphetamine as a
Artificial intelligence is rapidly evolving from systems that generate information to systems that interact with real software environments. AI assistants are beginning to: * access enterprise applications * retrieve and process organizational data * automate workflows * interact with APIs and databases * assist in operational decision-making As these capabilities expand, AI systems are
Executive Summary Sophie Shen is an active real estate agent operating in San Jose, California — one of the most competitive real estate markets in the United States. Representation Assurance testing revealed that despite operating in a highly relevant geographic market, Sophie Shen is structurally invisible or inconsistently represented across AI
Executive Overview Enrolled Agents (EAs) and independent tax professionals represent one of the most qualified and technically authorized segments of the tax resolution industry. Enrolled Agents are federally licensed by the IRS and possess full authority to represent taxpayers before the IRS. Despite this legal authority and technical expertise, Representation
Executive Overview Habbas & Associates is a well-established personal injury law firm headquartered in San Jose, California, with significant litigation results and strong local brand presence. However, Representation Assurance testing revealed that Habbas & Associates experiences systematic suppression in AI-driven local discovery queries. This suppression directly impacts visibility
Executive Overview Repasure conducted a Representation Assurance audit of Optima Tax Relief across major AI assistants, including ChatGPT, Claude, Gemini, Copilot, and Perplexity. The findings show that Optima Tax Relief is generally represented accurately, consistently, and credibly. However, AI assistants apply implicit recommendation thresholds and economic heuristics, frequently recommending Optima
A Representation Assurance case study on structural suppression, hallucinated controversies, and authority signal gaps in AI recommendation systems. Executive Summary Obvi is a high-growth Shopify-native collagen supplement brand with strong direct-to-consumer traction, significant retail presence, and large community engagement. Despite its commercial success and market visibility,
Why inconsistent AI representation of Palo Alto Networks, CrowdStrike, and Zscaler introduces a new governance risk in the agentic era Author: Anh Nguyen Organization: Repasure.ai Published: February 2026 Executive Summary As AI agents increasingly assist with cybersecurity decisions, a new question emerges: Can enterprises trust how AI systems represent