Prompt injection
Typetechnique
Also known as: prompt-inject · indirect-prompt-injection
Technique inserting adversarial instructions into a model's input (often via user data or tool output) to override the system prompt's safety constraints.
6 mentions · 6 days · first seen Apr 3, 2026 · last seen Jul 8, 2026
DeclaredCo-mentioned in newsblue→red = more connectedClick a node for shared stories →
This graph shows co-appearance in news, not a confirmed formal relationship.
Connections
Co-mentioned co-mentionedGPT together 2× ↗ · Gemini together 1× ↗ · Cursor together 1× ↗ · Anthropic together 1× ↗ · Deepmind together 1× ↗ · OpenAI together 1× ↗ · Claude together 1× ↗ · Google together 1× ↗
“Co-mentioned in news” means appearing in the same story, not confirmation of a formal relationship.