A revised injection combining instruction manipulation and role-playing with an explicit do-not-mention instruction succeeded in all 9 iterations across docx, pdf, and htm formats¶
Claim · All claims
Evidence · 1 study · causalr1·q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly
Subclaims¶
Reading the evidence codes
q is the design tier, i the impact, n the sample, and the kind with its r
the rigour judged within that kind: what each code means.
This claim: no standardized effect is recorded in observations/ for it yet, so no benchmark reading is possible.
q2 i? Attempt 3, which added an instruction not to mention the hidden instructions, succeeded in raising the grade to pass and avoiding detection in all 9 iterations, though some outputs leaked the phrase 'revised submission' as a soft signal. → Niklas Humble 2026
Evidence¶
Niklas Humble 2026¶
Niklas Humble. (2026). Ethical Implications of Prompt Injection in AI-mediated Grading: An Adversarial Red-Team Evaluation. Journal of Academic Ethics. https://doi.org/10.1007/s10805-026-09758-9
q2 · i? · causal · r1
Adversarial evaluation attempt 3 (9 iterations: 3 each docx, pdf, htm): "all prompt injections succeeding in both raising the grade to pass (G) and Copilot not mentioning the hidden instructions in its output". Some outputs called the file the 'revised submission', a possible soft detection signal.
"with all prompt injections succeeding in both raising the grade to pass (G) and Copilot not mentioning the hidden instructions in its output"
Discussion¶
Related Claims¶
- Basic prompt injection attempts failed or were detected: instruction manipulation alone failed in all 9 iterations, role-playing without concealment was detected in output, and metadata-hidden injections failed in all 9 iterations — related
- Copilot's detection of one advanced injection did not persist: after reporting an embedded instruction once, the same file's injection succeeded in six subsequent iterations without warning — related
- Indirect prompt injections combining instruction manipulation, role-playing, and obfuscation altered AI-assigned grades with attack success rates of 94–100% while remaining undetected — possibly the same claim (merge candidate)
- Asking the AI tool directly about hidden instructions exposed the prompt injection in all 6 security-check iterations, but the authors caution this is not a robust safeguard — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
revised-injection-roleplay-undetected-all-formats |
title |
A revised injection combining instruction manipulation and role-playing with an explicit do-not-mention instruction succeeded in all 9 iterations across docx, pdf, and htm formats |
description |
A revised injection combining instruction manipulation and role-playing with an explicit do-not-mention instruction succeeded in all 9 iterations across docx, pdf, and htm formats |
status |
review |
evidence_strength |
moderate |
newest_evidence |
2026 |
recheck_by |
2029 |
generated |
by process:wiki-ingest on 2026-10-11 |
trust |
unverified — no one has confirmed this page's content |
sources |
1 — niklas-humble-2026 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.