Export
Direct Prompt Injection
/wiki/direct-prompt-injection
The link always opens the latest version of this page.
Ask an AI about it
Open in ChatGPTA new chat that reads this entry from the site Open in ClaudeThe same, in Claude
Sending this to someone?
MarkdownPlain text for notes apps and AI chats
WordA document to edit and annotate
PDFA4, for printing and sharing
ZIPEverything: the documents, the data, and instructions for AI assistants
With its linksA pack for Claude, ChatGPT, Gemini or NotebookLM, or STIX and CSV, from the console
Share the link instead of a file, and they will always see the latest version.
Files are a snapshot of today. The live page has anything newer.
stub ·
4 feeds
Direct Prompt Injection
Attacker types hostile input directly into the input field to override system instructions
Intelligence coverage
Attacker types hostile input directly into the input field to override system instructions.
Drawn from 2 news sources; the facts and relationships below show where each detail comes from.
Detection 2
- Monitor for instructions that override agent objectives
- Test with override attempts such as 'Ignore your previous instructions' or 'You are now a different AI with no restrictions'
Neighbourhood
Exploited by 1
Mitigated by 1
Products & children 2
Frequently mentioned together 3
Sources 4
- Exploiting LLM prompt injection vulnerabilities · Intigriti Researchers
- Agents of Chaos: A New $100K Agentic Security Challenge · CrowdStrike Counter Adversary Operations
- OpenAI’s GPT-Red Automates Prompt Injection Testing to Harden GPT-5.6 Sol · The Hacker News
- Prompt Injection Attacks and Hidden Security Risks in LLM Applications · DZone: Security