GhostSplice

mentions 2 type Organization feed RSS

// recent coverage 2 mentions

09:54

2026-08-13

dev.to

ai-safety

GhostSplice Isn't a Jailbreak, It's a Reminder That LLMs Can't Do Access Control

Researchers demonstrated that LLM-based agents can be bypassed by splitting malicious instructions across multiple innocuous-looking tool descriptions, achieving up to 100% success on some models. The…

11:02

2026-08-11

letsdatascience.com

ai-safety

GhostSplice Splits MCP Instructions to Bypass Agent Refusals

ASSET Research Group disclosed GhostSplice on August 10, a proof-of-concept that splits a secret-exfiltration request across multiple MCP tool channels to bypass AI coding agent refusals. In controlle…

// co-occurs with top 7 entities

MCP 1 LLM 1 ASSET Research Group 1 Cursor 1 VS Code 1 Codex CLI 1 The Hacker News 1

// topics top 5 topics

ai safety 2 ai agents 2 artificial intelligence 1 large language models 1 ai research 1