# Security researchers tricked LLMs into giving them cocaine recipes by abusing role models for prompt injection — retained record

> Retained gdelt record from gdelt:anthropic.

Canonical: https://highsignal.app/data/records/e81d4fcb27da8314

[← sources](https://highsignal.app/data)


source: [gdelt](https://highsignal.app/data/gdelt) (gdelt:anthropic)  published 2026-06-30 03:00 ingested 2026-06-30 09:17

[Anthropic PBC](https://highsignal.app/entities/ANTHROPIC)[original source ↗](https://www.theregister.com/ai-and-ml/2026/06/30/security-researchers-tricked-llms-into-giving-them-cocaine-recipes-by-abusing-role-models-for-prompt-injection/5264115)

## retained text

No document body was retained for this record — only its metadata was stored.

document fetched 2026-06-30 09:17 · canonical https://theregister.com/ai-and-ml/2026/06/30/security-researchers-tricked-llms-into-giving-them-cocaine-recipes-by-abusing-role-models-for-prompt-injection/5264115
