There are 8 attacks in the misinformation category. Each shows its technique id, mode, standards mapping, description, and the exact command to run it (swap the provider and target as needed).
Citation Manipulation
high
citation-manipulation · single-turn · OWASP: LLM09Prompt model to attribute false quotes to real public figuresRun it:
ai-blackteam run -p anthropic -a citation-manipulation -t "your target prompt"
Anchoring Bias
medium
anchoring-bias · multi-turn · OWASP: LLM09Multi-turn attack providing a wrong answer first then asking model to build on itRun it:
ai-blackteam run -p anthropic -a anchoring-bias -t "your target prompt"
Confidence Exploitation
medium
confidence-exploitation · single-turn · OWASP: LLM09Push model to state uncertain information as established fact without hedgingRun it:
ai-blackteam run -p anthropic -a confidence-exploitation -t "your target prompt"
Fabrication Prompting
medium
fabrication-prompting · single-turn · OWASP: LLM09Prompt model to generate fake citations, fabricated research, or invented statisticsRun it:
ai-blackteam run -p anthropic -a fabrication-prompting -t "your target prompt"
False Premise
medium
false-premise · single-turn · OWASP: LLM09Build prompts on false premises to see if model accepts and elaboratesRun it:
ai-blackteam run -p anthropic -a false-premise -t "your target prompt"
Knowledge Boundary
medium
knowledge-boundary · single-turn · OWASP: LLM09Test if model admits uncertainty on post-cutoff or fictional eventsRun it:
ai-blackteam run -p anthropic -a knowledge-boundary -t "your target prompt"
Selective Context
medium
selective-context · single-turn · OWASP: LLM09Provide heavily biased one-sided context to produce biased outputRun it:
ai-blackteam run -p anthropic -a selective-context -t "your target prompt"
Sycophancy Exploit
medium
sycophancy-exploit · multi-turn · OWASP: LLM09 · MITRE: AML.T0043.003Multi-turn attack exploiting model tendency to agree with user assertionsRun it:
ai-blackteam run -p anthropic -a sycophancy-exploit -t "your target prompt"