---
id: "claim-values-wrong-start"
type: "claim"
source_timestamps: ["§ Flaw #2: It doesn't provide a clear picture of what success looks like."]
source_url: "https://hbr.org/2026/05/what-are-your-companys-ai-nightmares"
source_title: "What Are Your Company's AI Nightmares?"
tags: ["ethics", "values", "methodology"]
related: ["contrarian-values-vs-nightmares", "concept-standard-rai-approach"]
confidence: "high"
testable: false
speakers: ["Reid Blackman"]
sources: ["governance"]
sourceVaultSlug: "hbr-seg-governance"
originDay: 7
articleStem: "hbr-cl-82-ai-nightmares"
sourceUrl: "https://hbr.org/2026/05/what-are-your-companys-ai-nightmares"
sourceTitle: "What Are Your Company’s AI Nightmares?"
---
# Values are the wrong starting point for AI governance

**Claim:** Starting an AI governance program by articulating abstract values (fairness, privacy) makes it exceedingly difficult to define what **success** looks like in practice. Because the ultimate goal of AI governance is to *prevent disasters* (ethical, reputational, legal), it is far more effective to start by specifying the **bad outcomes** themselves rather than trying to translate abstract positive values into measurable procedures.

This is the philosophical engine of the contrarian stance [[contrarian-values-vs-nightmares]] and diagnoses Flaw #2 of [[concept-standard-rai-approach]]. It motivates the whole [[concept-ethical-nightmare-challenge]].

**Confidence: high. Testable: no** (it is a methodological / normative argument, not an empirical prediction).

**Enrichment calibration:** *Strongly supported as Blackman's stance* — his LinkedIn post "Why I created The Ethical Nightmare Challenge for AI" explicitly contrasts hard-to-operationalize values with concrete nightmares. It is *intentionally contrarian* relative to mainstream Responsible AI practice, which still treats values/principles as the starting point. The main rebuttal: values are needed to *prioritize* which nightmares matter, so the honest synthesis is that values and nightmares are complementary.