---
id: "action-reframe-overrides"
type: "action-item"
source_timestamps: ["§ The Takeaways for Managers", "¶16"]
tags: ["security-policy", "incident-response", "prompt-injection"]
related: ["claim-overrides-signal-design-flaws", "contrarian-overrides-not-malicious"]
action: "Investigate the AI's interaction design before punishing employees who attempt to bypass or override the system."
outcome: "Resolves the root cause of user resistance (bad design) rather than treating the symptom (overrides)."
speakers: ["Aleksandra Przegalinska", "Tamilla Triantoro", "Leon Ciechanowski", "Konrad Sowa", "Anna Kovbasiuk", "Richard B. Freeman"]
sources: ["tail1"]
sourceVaultSlug: "hbr-seg-tail1"
originDay: 1
articleStem: "hbr-tail-113-ai-personality-problem"
sourceUrl: "https://hbr.org/2026/06/does-your-ai-have-a-personality-problem"
sourceTitle: "Does Your AI Have a Personality Problem?"
---
# Reframe Prompt Injection as UX Feedback

**Action:** Investigate the AI's interaction design *before* punishing employees who attempt to bypass or override the system.

**Outcome:** Resolves the root cause of user resistance (bad design) rather than treating the symptom (overrides).

**Detail:** When ordinary employees attempt to bypass AI guardrails or use prompt injection to change the AI's character, do **not** immediately assume malicious intent or crack down with new restrictions and monitoring. Instead, treat these events as **diagnostic signals** that the AI's current persona is provoking resistance (as established in [[claim-overrides-signal-design-flaws]] and reframed in [[contrarian-overrides-not-malicious]]). Investigate whether the [[concept-dark-triad-ai|interaction design]] is overly rigid, unhelpful, or hostile.

This is Step 3 of the [[framework-managerial-takeaways|three-step governance framework]]. *Balance caveat (from enrichment):* security practitioners argue overrides remain a genuine risk surface — the mature stance treats them as **both** a usability signal *and* a security concern.