CVE-2026-82533

CVE

Threat entity extracted from intelligence sources

Frequency

occurrences

First Seen

September 9, 2026

Last Seen

September 9, 2026

API

Exploited in Wild

—

Ransomware Use

—

Public Exploits

—

Attack Vector

—

Vulnerability Overview

Exploitation Activity

Exploitation Intelligence

Critical Flaw in DeepSeek Harness Allows AI Agents to Escape Sandbox70.5

A critical vulnerability (CVE-2026-82533) was discovered in DeepSeek Harness, an open-source tool for AI coding agents, enabling a sandboxed agent to disable its own confinement using a single shell command. The flaw stems from the tool's unauthenticated API, which relied on the client-supplied 'Hos...