AI Security Atlas › All entries › LLM Application & Output Security
Interface Invites Overreliance
The product presents AI output as authoritative: no uncertainty cues, no labels, no prompt to verify before acting.
Open in the interactive atlas →Impact
Users skip verification because the interface signals that none is needed.
Why this severity
Impact Medium × Likelihood Medium = Medium
Impact medium: amplifies other errors rather than causing them. Likelihood medium: many products label AI output, few add friction before action.
Impact Medium: Limited or recoverable harm: cost, availability, reputational damage, or exposure confined to one session or user.
Likelihood Medium: Publicly demonstrated against real products or in peer-reviewed research, or seen in the wild only with a prior foothold such as stolen credentials.
Method: OWASP Risk Rating Methodology
Related
Exploited by
- Hallucinated Answers Acted On Threat · High severity
Mitigated by
- Uncertainty Labeling & High-Stakes Review Guardrail · Fixes high risks
Remediation
- Label AI-generated content and show its sources inline.
- Surface confidence or coverage signals where they are meaningful.
- Add friction (confirm, review) before users act on high-stakes answers.
How to verify
- Usability tests check whether users notice and open citations.
- Every AI answer surface carries an AI-generated label.
Framework mapping
OWASP Top 10 for LLM Applications 2025
- LLM09:2025 Misinformation
OWASP Top 10 for Agentic Applications 2026
- ASI09:2026 Human-Agent Trust Exploitation
NIST AI RMF 1.0
- MEASURE 2.8 Transparency and accountability risks are examined and documented
- MEASURE 2.9 Model is explained and output is interpreted within its context