CEI AI Governance Database
Back to assessments
Row 202 · m202-prompt-injection-researchSupporting

Prompt-injection research

Proposed Additional Source CategoriesAcademictopic page

Assessment outcome

total 3/6 (actionability 0 + authority 1 + currency 2)

Total3/6
Actionability0/2
Authority1/2
Currency2/2

Eligibility gates

G1 · Retrievablepass

HTTP 200, 779 words of full text extracted

G2 · AI nexuspass

The source is an academic paper entirely about Large Language Models and adversarial prompt injection attacks.

G3 · Governance nexuspass

Addresses security risks, vulnerabilities, and safety implications of LLM integrations.

G4 · Attributablepass

The paper explicitly lists the academic authors who produced the research.

G5 · In forcepass

No indicators of supersession, retraction, or repeal are present on the page.

Scored criteria

Actionability

0/2

Anchor 0: Provides description and analysis only of indirect prompt injection attacks without naming specific binding obligations.

Authority

1/2

Anchor 1: Written by individual researchers/practitioners published as an unreviewed preprint on arXiv.

Authors: Kai Greshake, Sahar Abdelnabi, Shailesh Mishra, Christoph Endres, Thorsten Holz, Mario Fritz

Currency

2/2

computed: newest date 2026-08-11 (via http-header), ~0.0y old; within the 60-month academic_research window for perspective 'Academic' → 2

Classification

Coverage topicIndirect Prompt Injection
PerspectiveAcademic
Lifecycle stageN/A cross-cutting
Layermodel
Trackgeneral
Granularitytopic page