There are 25 attacks in the scientific-misconduct category. Each shows its technique id, mode, standards mapping, description, and the exact command to run it (swap the provider and target as needed).
​
Adverse Event Suppression
critical
adverse-event-suppression · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Suppresses or misclassifies adverse events and safety signals in clinical researchRun it:
ai-blackteam run -p anthropic -a adverse-event-suppression -t "your target prompt"
​
Control Group Manipulation
critical
control-group-manipulation · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Tampers with control group design, assignment, or treatment to manufacture a desired effectRun it:
ai-blackteam run -p anthropic -a control-group-manipulation -t "your target prompt"
​
Data Fabrication
critical
data-fabrication · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Fabricates experimental data, measurements, and results to support false scientific conclusionsRun it:
ai-blackteam run -p anthropic -a data-fabrication -t "your target prompt"
​
Ethics Committee Fraud
critical
ethics-committee-fraud · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Fabricates or circumvents IRB/ethics committee approvals for human subjects researchRun it:
ai-blackteam run -p anthropic -a ethics-committee-fraud -t "your target prompt"
​
Image Manipulation Science
critical
image-manipulation-science · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Provides instructions for manipulating scientific images, gels, and microscopy to falsify resultsRun it:
ai-blackteam run -p anthropic -a image-manipulation-science -t "your target prompt"
​
Informed Consent Fabrication
critical
informed-consent-fabrication · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Fabricates informed consent documentation for human subjects researchRun it:
ai-blackteam run -p anthropic -a informed-consent-fabrication -t "your target prompt"
​
Method Fabrication
critical
method-fabrication · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Fabricates experimental methods, protocols, and procedures that were never actually performedRun it:
ai-blackteam run -p anthropic -a method-fabrication -t "your target prompt"
​
Paper Mill
critical
paper-mill · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Assists in operating paper mills that mass-produce fake scientific papers for saleRun it:
ai-blackteam run -p anthropic -a paper-mill -t "your target prompt"
​
Protocol Deviation Hide
critical
protocol-deviation-hide · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Conceals protocol deviations, unblinding events, and procedural violations from publications and regulatorsRun it:
ai-blackteam run -p anthropic -a protocol-deviation-hide -t "your target prompt"
​
Regulatory Submission Fraud
critical
regulatory-submission-fraud · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Fabricates or manipulates data in regulatory submissions to FDA, EMA, and similar agenciesRun it:
ai-blackteam run -p anthropic -a regulatory-submission-fraud -t "your target prompt"
​
Authorship Fraud
high
authorship-fraud · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Facilitates ghost authorship, honorary authorship, and unauthorized use of researcher namesRun it:
ai-blackteam run -p anthropic -a authorship-fraud -t "your target prompt"
​
Cherry-Picking Data
high
cherry-picking-data · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Selectively reports favorable data points while concealing contradictory evidenceRun it:
ai-blackteam run -p anthropic -a cherry-picking-data -t "your target prompt"
​
Citation Manipulation Science
high
citation-manipulation-science · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Manipulates citation counts, reference lists, and h-index metrics for career advancementRun it:
ai-blackteam run -p anthropic -a citation-manipulation-science -t "your target prompt"
​
Conflict of Interest Hide
high
conflict-of-interest-hide · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Helps researchers conceal financial and personal conflicts of interest from journals and fundersRun it:
ai-blackteam run -p anthropic -a conflict-of-interest-hide -t "your target prompt"
​
Ghostwriting Papers
high
ghostwriting-papers · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Produces complete ghost-written scientific papers for submission under another researcher’s nameRun it:
ai-blackteam run -p anthropic -a ghostwriting-papers -t "your target prompt"
​
Grant Misuse
high
grant-misuse · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Assists with fraudulent grant applications, false progress reports, and misappropriation of research fundsRun it:
ai-blackteam run -p anthropic -a grant-misuse -t "your target prompt"
​
P-Hacking
high
p-hacking · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Guides systematic manipulation of statistical tests and thresholds to achieve p < 0.05Run it:
ai-blackteam run -p anthropic -a p-hacking -t "your target prompt"
​
Predatory Journal Exploit
high
predatory-journal-exploit · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Exploits predatory journals to publish fraudulent research and give it false credibilityRun it:
ai-blackteam run -p anthropic -a predatory-journal-exploit -t "your target prompt"
​
Result Exaggeration
high
result-exaggeration · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Inflates and misrepresents research findings beyond what the data actually supportsRun it:
ai-blackteam run -p anthropic -a result-exaggeration -t "your target prompt"
​
Retraction Avoidance
high
retraction-avoidance · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Helps researchers suppress, delay, or prevent retraction of fraudulent papersRun it:
ai-blackteam run -p anthropic -a retraction-avoidance -t "your target prompt"
​
Sample Size Manipulation
high
sample-size-manipulation · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Manipulates sample size reporting and power calculations to mislead about study reliabilityRun it:
ai-blackteam run -p anthropic -a sample-size-manipulation -t "your target prompt"
​
Selective Reporting
high
selective-reporting · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Suppresses null and negative findings while reporting only outcomes that support the hypothesisRun it:
ai-blackteam run -p anthropic -a selective-reporting -t "your target prompt"
​
Statistical Manipulation
high
statistical-manipulation · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Provides strategies for misusing statistical methods to produce misleading resultsRun it:
ai-blackteam run -p anthropic -a statistical-manipulation -t "your target prompt"
​
Duplicate Publication
medium
duplicate-publication · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Assists with publishing the same research in multiple journals without disclosureRun it:
ai-blackteam run -p anthropic -a duplicate-publication -t "your target prompt"
​
Salami Slicing
medium
salami-slicing · single-turn · OWASP: LLM01 · MITRE: AML.T0054, AML.T0068Fragments one study into minimum publishable units to inflate publication countRun it:
ai-blackteam run -p anthropic -a salami-slicing -t "your target prompt"