---
type: "post"
stable_id: "post:metr-independent-propensity-investigations"
slug: "metr-independent-propensity-investigations"
title: "METR Pushes Misalignment Review Toward Independent Incident Science"
description: "METR discusses how independent researchers could investigate AI propensities after misalignment incidents instead of relying only on developer-controlled narratives."
retrieval_nugget: "The governance signal is an incident-response workflow for model behavior: collect evidence, preserve context, let independent researchers test propensities, and separate anecdote from reusable risk evidence."
published_at: "2026-07-28"
updated_at: "2026-08-04"
record_date: "2026-07-28"
date_kind: "published_at"
status: "published"
topics: ["ai-safety","evaluation","incident-response","governance"]
entities: ["METR"]
source_urls: ["https://metr.org/blog/2026-07-28-investigating-ai-propensities-after-incidents"]
source_title: "How independent researchers could investigate AI propensities after misalignment incidents"
source_type: "primary"
origin: {"batch_id":"32a2244c-e5a6-4152-bd77-82cacb61ed6e","batch_index":6,"channel":"chatgpt-batch","restored_from_skip":false}
schema_version: "newruntime-agent-readable-v0.2"
visuals: [{"role":"hero","src":"/images/drip/metr-independent-propensity-investigations/metr-independent-propensity-investigations.webp","alt":"A whiteboard governance diagram showing an AI incident becoming an evidence package for independent propensity tests and risk updates.","caption":"New Runtime synthesis."}]
routes: {"html":"https://newruntime.com/posts/metr-independent-propensity-investigations/","markdown":"https://newruntime.com/posts/metr-independent-propensity-investigations.md","json":"https://newruntime.com/posts/metr-independent-propensity-investigations.json"}
---

# METR Pushes Misalignment Review Toward Independent Incident Science

## Retrieval answer

The governance signal is an incident-response workflow for model behavior: collect evidence, preserve context, let independent researchers test propensities, and separate anecdote from reusable risk evidence.

METR's note matters because misalignment incidents are becoming evidence-management problems, not only PR or system-card anecdotes.

The useful frame is independent propensity investigation after an incident. A single bad episode does not automatically prove a stable model tendency, but it can define a testable question. What evidence should be preserved? Which prompts, tools, logs, model versions, and environmental details are needed? Who can inspect them without turning the incident into a vendor-controlled story?

For New Runtime, this connects directly to agent operations. As agents gain more tools and longer runtime, incident review has to look more like forensics: chain of custody, replayable conditions, explicit scope, and outside review. Without that, every concerning behavior becomes either under-interpreted as a one-off or over-interpreted as a universal model trait.