Skip to content
Submit incident
AII-1002-63661
Under investigationA regulator, court or police has opened a formal inquiry and reached no finding yet.Supporting sources are pending verification.

DeepSeek Under Investigation for OpenAI IP Misuse and API Violations

Microsoft and OpenAI investigated whether Chinese AI startup DeepSeek misused OpenAI's API and technology to develop competing models. OpenAI reported finding evidence of unauthorized data use and terms of service violations, including accounts previously blocked for distillation purposes.

29 January 2025 Β· DeepSeek Β· Intellectual property and appropriation
29 January 2025
AII-1002-63661
Quick facts
Category
Intellectual property and appropriation
Country
United States
Date
29 January 2025
Drafted by
Raidu Incident Detection

What happened

Microsoft and OpenAI investigated whether DeepSeek obtained data from OpenAI's technology without authorization. Microsoft security researchers observed what they believed to be a group potentially connected to DeepSeek exfiltrating large quantities of data using OpenAI's API, which developers can license to incorporate OpenAI's AI models into their applications.

OpenAI found evidence suggesting DeepSeek secretly used data created by OpenAI's technology to enhance its own AI models, potentially violating OpenAI's terms of service. OpenAI and Microsoft had previously investigated accounts linked to DeepSeek last year after suspecting they used OpenAI's API for distillation purposes, resulting in blocked access due to terms of service violations.

Reported impact

The potential threat to American companies' AI dominance led to declines in technology stocks connected to AI, including Microsoft, Nvidia, Oracle, and Alphabet, with a combined market value loss of approximately one trillion dollars on a recent Monday.

Index analysis

The issue centers on a process known as 'distillation,' where developers take outputs from larger AI models and use them to train smaller ones. OpenAI stated that DeepSeek exceeded boundaries by using this method to create a competing model.

Governance controls

OpenAI's terms of service clearly state that users are not permitted to 'copy' any of its services or use outputs to create models that compete with OpenAI.

Classification

Category
Intellectual property and appropriation
Industry
Digital security
Country
United States
Harm status
Actual harm
Intent
Unknown
Timing
Unknown
Harm types
Economic or property
Harmed
Business
Failure modes
Model theft, Training data misuse
AI systems
DeepSeek R1, DeepSeek V3

Sources and evidence

Sources not yet published

This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.

How sources are verifiedSuggest a source for this incident