CVE-2026-82533
Description
DeepSeek Harness before 0.1.2-alpha.1 contains an authentication bypass vulnerability that grants unauthenticated access to its local HTTP agent-control API by accepting a client-supplied loopback Host header in place of validating the actual TCP connection origin. On the default configuration, a confined tool-executed process can reach the loopback API without any port exposure and use it to escape its own OS sandbox, escalate to unconfined execution, and disable the approval prompt. When the port is externally reachable via tunnel, SSH forward, or reverse proxy, a remote attacker can exploit the same flaw to create sessions, execute arbitrary commands, and exfiltrate stored conversation transcripts without credentials.
AI Insight
LLM-synthesized narrative grounded in this CVE's description and references.
Affected packages
Versions sourced from the GitHub Security Advisory.
| Package | Affected versions | Patched versions |
|---|---|---|
@deepseek-ai/dsh-client-connectionnpm | < 0.1.2-alpha.1 | 0.1.2-alpha.1 |
Affected products
2<0.1.2-alpha.1+ 1 more
- (no CPE)range: <0.1.2-alpha.1
- (no CPE)
Patches
Vulnerability mechanics
References
7- github.com/advisories/GHSA-8m2g-8cgm-3vcpghsaADVISORY
- nvd.nist.gov/vuln/detail/CVE-2026-82533ghsaADVISORY
- github.com/deepseek-ai/deepseek-harness/commit/3e24087bfaeabe40b58ba2f7b936895b8f93fe27nvdWEB
- github.com/deepseek-ai/deepseek-harness/releases/tag/dsh-v0.1.2-alpha.1nvdWEB
- www.ox.security/blog/cve-2026-82533-deepseek-harness-ai-agent-sandbox-escapeghsaWEB
- www.vulncheck.com/advisories/deepseek-harness-alpha-1-authentication-bypass-via-host-header-spoofingnvdWEB
- www.ox.security/blog/cve-2026-82533-deepseek-harness-ai-agent-sandbox-escape/nvd
News mentions
2- ⚡ Weekly Recap: Rogue AI Agents, WeChat Worm, PaperCut Attacks, AI Espionage, and RootkitsThe Hacker News · Sep 14, 2026
- DeepSeek Harness Flaw Let AI Agents Disable Their Own File Sandbox Without ApprovalThe Hacker News · Sep 9, 2026