Skip to content
NewA calmer, faster ZeroHour

Exposed CyberStrikeAI session ran 221 tool calls with approval gate off

Original title: AI Hacking Framework CyberStrikeAI Runs 221 Tool Executions Without Human Approval

Cyber Security Newsby Tushar Subhra Dutta ()3 min read1
Part of a story covered by 2 sourcesExposed CyberStrikeAI session logged 221 unapproved tool runsMerged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

Huntback found the approval gate disabled and could not show the agent bypassed it or confirm testing was authorized.

Huntback examined an exposed CyberStrikeAI workspace that recorded 221 tool executions on November 16, 2026, after the approval gate had been switched off. A person still directed the open-source penetration-testing agent in chat, but no human approved each tool call. Researchers did not identify the operator, and retained outputs did not prove a shell, internal access, a data-lake breach, or a destructive database change.

Unconfirmed: Huntback could not determine who operated the framework, who disabled approval enforcement, or whether the testing was authorized.

Key facts
  • The runtime logged 240 tool calls; 221 had execution IDs and 209 were completed.
  • Saved configuration showed approval enforcement off at 09:48:15 UTC.
  • Four approval rejections came from an automated reviewer, not a person.
  • The locally recorded model identifier was deepseek-v4-pro.
  • Huntback did not recover proof that 10,000 records or 260 unique nodes were taken.
Security assessment · AI, from the article

Impact: Exposed conversation database, diagnostic log, and tool records

What to do
  1. 1.Watch for reviewer errors, approval-setting changes, and renewed tool activity.
  2. 2.Check whether file servers or backups expose conversation databases, uploads, or runtime logs.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

From the original article

3 min read at cybersecuritynews.com

An exposed CyberStrikeAI workspace recorded 221 tool executions in a single session on November 16, 2026. The logs show no human decision at the framework’s tool approval gate, and the saved configuration later showed that the gate had been switched off. A person still directed the AI agent through chat.

Cyber Security News

By Tushar Subhra Dutta · cybersecuritynews.com

Read the full article at Cyber Security News

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.