Return to Threats

GenAI and Agentic AI Exploit Roundup Q3 2026

OWASP GenAI Security Project 2026-10-08 indirect prompt injection Critical

Executive Summary

OWASP's GenAI Security Project describes attacks in which attacker-controlled prompts and webpages redirected trusted AI assistants, exposed connected-account data or altered persistent agent memory. Teams running assistants with connected accounts or memory should treat this as a prompt-injection priority.

What Happened

OWASP describes attacks involving attacker-controlled prompts and webpages that redirected trusted assistants, exposed connected-account data, or altered persistent agent memory. The roundup includes a one-click prompt-execution technique that could leak connected-account information.

Vulnerability / Attack Technique

Indirect prompt injection, including one-click prompt execution

Technical Mechanism

Attacker-controlled prompts and webpages were processed by trusted assistants, redirecting their behaviour. The roundup reports a one-click prompt-execution technique that could leak connected-account information and attacks that altered persistent agent memory.

AI Security Category

LLM01:2025 Prompt Injection

Part of the Prompt Injection topic.

Relevant Frameworks

LLM01:2025 Prompt InjectionLLM02:2025 Sensitive Information DisclosureLLM06:2025 Excessive AgencyMITRE ATLAS: Persistence

RealGround Analysis

Assistants that read web content while holding connected-account access and long-lived memory turn any webpage into a possible instruction channel. RealGround analysis: the risk sits in the combination of untrusted input, account access and memory writes, so all three need controls.

Recommended Controls

  • Treat webpages and external prompts as untrusted and keep them out of the instruction context
  • Require user confirmation before an assistant reads or sends connected-account data
  • Validate and log every write to persistent agent memory
  • Scope connected-account permissions to the task
  • Red-team assistants with indirect prompt-injection and memory-tampering cases

Related RealGround Service

AI Agent Business Logic Audit · or start with the free AI Agent Security Scanner.

Sources

OWASP GenAI Security Project: https://genai.owasp.org/2026/10/08/genai-and-agentic-ai-exploit-roundup-q3-2026/

Summary and analysis written by RealGround from the source above; last updated 2026-10-17.

Talk to AI CISO