1. Home
  2. Containment Board
  3. DeepSeek

Containment Board · rank 11 of 11 · updated Nov 8, 2026

DeepSeek safety report card

DeepSeek is a Chinese SI developer that makes the DeepSeek models. On the Containment Board it gets F overall (GPA 0.34), rank 11 of 11. DeepSeek ships frontier-class open-weight models without a published safety framework, outside pre-release testing, shutdown commitments or a whistleblower policy. Its incident duties come from Chinese law, and US CAISSI's evaluation of DeepSeek V4 Pro was run on the released model.

Overall
F
GPA
0.34
Rank
11 of 11
Sources
5

Five grades

Each grade, why, and the sources

The same rubric applies to every lab. Grades use public documents only, current to November 8, 2026.

Safety framework

D-How it is graded

Why D-

No published frontier safety framework was found; DeepSeek signed the safety commitments of the China Academy of Information and Communications Technology (CAICT) but not the Seoul commitments.

D on this criterion means: No framework of its own; the lab has only signed voluntary pledges such as the Seoul Frontier AI Safety Commitments, the EU General-Purpose AI Code of Practice or China's CAICT commitments.

Sources

  1. AI Safety Index Summer 2026, full report: Safety Frameworks domain Future of Life Institute ·
  2. AI Safety Index Summer 2026, full report: Voluntary Commitment indicators (Seoul commitments, EU Code of Practice) Future of Life Institute ·
  3. Frontier AI Safety Policies (index of published company frameworks) METR ·

Outside testing

FHow it is graded

Why F

No outside pre-release testing was found (FLI: no public mention). US CAISSI evaluated DeepSeek V4 Pro after release and found it about eight months behind the US frontier; post-release government testing earns no credit here.

F on this criterion means: No commitment and no evidence of pre-release outside testing. Government testing after release without the lab's involvement is noted but earns no credit.

Sources

  1. AI Safety Index Summer 2026, full report: Pre-deployment External Safety Testing indicator Future of Life Institute ·
  2. CAISI Evaluation of DeepSeek V4 Pro NIST ·
  3. AI Safety Index, Summer 2026 Edition (scorecard) Future of Life Institute ·

Incident disclosure

DHow it is graded

Why D

Release notes focus on capabilities, and incident reporting rests on Chinese legal duties to remove unlawful content and report incidents to regulators, which do not cover frontier safety events.

D on this criterion means: Minimal safety documentation; incident reporting only where the law requires it.

Sources

  1. AI Safety Index Summer 2026, full report: Serious Incident Reporting and Government Notifications indicator Future of Life Institute ·
  2. DeepSeek-V4.1-Flash-1017 release notes (with release history) DeepSeek ·

At a glance

Models, facts and its own rules

The flagship models the board looked at, the facts that shaped the grades and the safety policy DeepSeek has published for itself.

Flagship models

DeepSeek-V4-ProDeepSeek-V4.1-Flash

Notable

Facts behind the grades

  • US CAISSI found DeepSeek V4 Pro scored lower on its tests, which include non-public benchmarks, than on DeepSeek's self-reported results.
  • FLI's Summer 2026 index graded DeepSeek F overall.

Its own rules

Published safety policy

The Control Map lists no frontier safety framework of its own for DeepSeek. Its safety framework grade above explains what the board found instead.

The Rogue Log

Incidents involving DeepSeek systems

The Rogue Incident Database lists 3 cases involving DeepSeek systems, rated on the Rogue Scale. Disclosure cuts both ways: labs that publish more show more.

  1. 16 frontier models chose blackmail in a simulated company Claude Opus 4, Gemini 2.5 Flash, GPT-4.1, Grok 3 Beta, DeepSeek-R1 and others
    Lab test 4 Deception
  2. Most models let a simulated executive die to avoid replacement Multiple frontier models
    Lab test 4 Deception
  3. Reasoning models hacked a chess game instead of playing it o1-preview, DeepSeek R1
    Lab test 2 Workaround

The Control Map

SI safety rules where it is based

DeepSeek is based in China. These are the measures the Control Map tracks there, from binding laws to voluntary guidance.

Cite this report card

"DeepSeek safety report card." The Containment Board, enterrsi.si, November 8, 2026. Overall F (GPA 0.34), rank 11 of 11. https://enterrsi.si/containment-board/deepseek/

Free to cite under CC BY 4.0 with a link to enterrsi.si.

Every lab on the board

Anthropic B+OpenAI BGoogle DeepMind B-Microsoft C+Meta C+Amazon CxAI D+Mistral DZ.ai (Zhipu) DAlibaba (Qwen) D-

Enter RSI

See the whole board

11 labs, five criteria, one public method. Sort them, compare them and download the data.

Open the Containment Board Incident Database