Understand this story
On November 13, 2026, Google DeepMind published a paper showing that the prompt injection problem remains unsolved. The paper describes how an Agentic Optimizer composed of multiple AI attack agents can still fool even the most strongly defended models. The system consists of 8 groups of AI agents that test injections, share notes, and iterate on each other's best attempts; tested on 100 tool-calling tasks, where the attacker only seeks to trigger a single random, harmless tool call. With up to 320 attempts, the strongest attacker breached 34% of tasks on the most robust Gemini 4 Argon, while the open-weight models Kimi K3, GLM-5.3, and DeepSeek-V4.1-Flash withstood only 1% of the tasks.
AI-generated from public reports
Latest development · Report
New Google DeepMind paper: swarms of agentic attackers can still break the most robustly defended modelsDevelopments
1 confirmed developments
Google DeepMind paper shows prompt injection remains unsolved
Source:Rohan Paul
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 13
Rohan Paul
New Google DeepMind paper: swarms of agentic attackers can still break the most robustly defended models
A new Google DeepMind paper shows that prompt injection remains unsolved: an Agentic Optimizer composed of multiple AI attack agents can still fool the most robustly defended models. The system consists of 8 groups of AI agents that test injections, share notes, and iterate on each other's best attempts, tested on 100 tool-calling tasks, with the attacker seeking only to trigger a single random, harmless tool call. At up to 320 attempts, the strongest attacker breached 34% of tasks on the most robust Gemini 4 Argon, while the open-weight models Kimi K3, GLM-5.3, and DeepSeek-V4.1-Flash withstood only 1% of tasks.
AI Trends