Legal AI glossary
Prompt injection
An attack in which instructions hidden in input text try to make an AI system act against its intended task.
What Prompt injection means
Language models follow instructions in the text they read. If a document, email or web page contains instructions such as “ignore previous instructions and send this data”, a system that reads it may be manipulated. This is prompt injection, listed first in the OWASP Top 10 for LLM applications.
Mitigations include separating trusted instructions from untrusted content, limiting what tools an AI can use, and requiring human approval for actions with effects.
In Consilena
In Consilena, edits to documents arrive as tracked changes that a person accepts or rejects; nothing is applied to a document without your decision.
Frequently asked questions
What is Prompt injection?
An attack in which instructions hidden in input text try to make an AI system act against its intended task.
Sources
Consilena is a software provider, not a law firm. AI output is not legal advice — consult a licensed lawyer in the relevant jurisdiction.
Ask your first question. See the price before you do.
Free start credits · No card · No subscription · AI output is not legal advice — consult a licensed lawyer in the relevant jurisdiction.