- Home
- Encyclopedia
- DeepSeek
RSI Encyclopedia · Lab
DeepSeek
DeepSeek is a Chinese SI developer that makes the DeepSeek models. Its Containment Board grade is F (GPA 0.34), rank 11 of 11.
- Overall grade
- F
- Board rank
- 11 of 11
- Models listed
- 3
- Rogue Log cases
- 3
The lab
DeepSeek is a Chinese SI developer that makes the DeepSeek models. 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.
Safety grades
On the Containment Board DeepSeek gets F overall (GPA 0.34), rank 11 of 11. Each grade, its reasons and its sources are on the DeepSeek report card.
- Safety frameworkD-
- Outside testingF
- Shutdown and pauseF
- Incident disclosureD
- Whistleblower protectionF
Notable facts
- 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.
Rules where it is based
- In force Generative AI Measures China
- In force Labeling Measures China
- In force Cybersecurity Law SI clause China
- Voluntary or guidance Framework 3.0 China
Sources
- AI Safety Index Summer 2026, full report: Whistleblowing Protection indicators and Reporting Culture and Whistleblowing Track Record Future of Life Institute ·
- Frontier AI Safety Policies (index of published company frameworks) METR ·
- CAISI Evaluation of DeepSeek V4 Pro NIST ·
- AI Safety Index, Summer 2026 Edition (scorecard) Future of Life Institute ·
- DeepSeek-V4.1-Flash-1017 release notes (with release history) DeepSeek ·
In the Rogue Log
Incidents involving DeepSeek systems
3 cases in the Rogue Incident Database involve DeepSeek systems. The most recent are shown here.
Jun 20, 2025 · Lab test
16 frontier models chose blackmail in a simulated company
Read the case4 DeceptionJun 20, 2025 · Lab test
Most models let a simulated executive die to avoid replacement
Read the case2 WorkaroundFeb 18, 2025 · Lab test
Reasoning models hacked a chess game instead of playing it
Read the caseThe RSI Encyclopedia
Keep reading
207 entries on the terms, labs, models, laws and events of rogue SI, plus every case in the Rogue Incident Database.
Enter RSI
Every term, lab, model and law
Browse the full A to Z index, filter by category or search for a name.