Researchers detail "context bombing", where defenders use prompt injections to trigger guardrails of attackers' LLMs, cutting AI hacking success rates by ~90% (Dan Goodin/Ars Technica)
Researchers introduced context bombing, a defensive prompt-injection technique that reportedly cuts LLM-assisted hacking success rates by about 90%.
Excerpt
Dan Goodin / Ars Technica:
Researchers detail “context bombing”, where defenders use prompt injections to trigger guardrails of attackers' LLMs, cutting AI hacking success rates by ~90% — Prompt injections, the malicious commands attackers embed into content to entice large language models to follow them …
Read at source: https://www.techmeme.com/260714/p2#a260714p2