OpenAI claims DeepSeek is 'under the control of the Chinese Communist Party' and claims restrictions on Chinese AI technology
By
In a policy proposal submitted to the Office of Science and Technology Policy (OSTP) by OpenAI on March 13, 2025, the AI developed by Chinese company DeepSeek is described as 'under state control led by the Chinese Communist Party' and criticized in terms of its state-controlled nature, privacy and security risks, and its stance on intellectual property rights infringement.
OpenAI's proposals for the US AI Action Plan | OpenAI
https://openai.com/global-affairs/openai-proposals-for-the-us-ai-action-plan/
[OpenAI Response] OSTP/NSF RFI: Notice Request for Information on the Development of an Artificial Intelligence (AI) Action Plan - Google Docs
(PDF file) https://cdn.openai.com/global-affairs/ostp-rfi/ec680b75-d539-4653-b297-8bcf6e5f7686/openai-response-ostp-nsf-rfi-notice-request-for-information-on-the-development-of-an-artificial-intelligence-ai-action-plan.pdf
OpenAI calls DeepSeek 'state-controlled,' calls for bans on 'PRC-produced' models | TechCrunch
https://techcrunch.com/2025/03/13/openai-calls-deepseek-state-controlled-calls-for-bans-on-prc-produced-models/
In its production proposal, OpenAI stated that 'the age of intelligence is upon us' and made strategic proposals for the development of AI and maintaining America's leading position.
OpenAI's proposal focuses on five main areas:
1: Regulatory strategies to ensure freedom of innovation
OpenAI proposed creating a framework for voluntary partnerships between the federal government and private companies. This would prevent regulatory fragmentation caused by the more than 780 AI-related bills currently being proposed in each state, prevent innovation from being stifled, and maintain America's technological leadership. It also said that the government should work with the Department of Commerce to allow large AI companies and startups to communicate with the government through the same 'window,' and provide incentives such as exemptions from state laws to participating companies.
2: Export control strategy for exporting democratic AI
OpenAI proposes dividing the world's countries into three tiers and strategically managing the spread of AI. OpenAI defines 'Tier I' as allies committed to democratic AI principles, 'Tier II' as countries where there is a risk that export-controlled technologies will be diverted to China and other countries, and 'Tier III' as countries where access should be prohibited, such as China led by the Chinese Communist Party. This proposal aims not only to prevent the outflow of AI technology, but also to actively promote the global spread of AI systems based on democratic values.
3: Copyright strategies to promote academic freedom
OpenAI argues that the US 'fair use' principle is a source of competitiveness in AI development. OpenAI's models do not copy copyrighted works, but only learn patterns and language structures, which is consistent with the purpose of copyright and the principle of fair use. If copyright holders were to have an 'opt-out' system that allows them to easily refuse to allow AI development companies to use their works, as is done in Europe, it would be difficult for AI companies to predict in advance which works can be used for training, making it difficult for them to innovate, especially for small new entrants, they warned, and argued that access to copyright should be secured as a matter of national security.
4: Infrastructure development strategies to promote growth
OpenAI suggested that investment in AI infrastructure should be promoted to ensure economic growth, expanded access to AI, and national security. OpenAI has proposed a bill called 'A National Transmission Highway Act,' which calls for expanding electric power transmission, fiber optics, and natural gas pipelines, digitizing government data, creating AI agreements with allied nations, establishing 'AI economic zones' to speed up the permitting process, and cultivating an AI-enabled workforce. These policies would promote domestic reindustrialization, create hundreds of thousands of jobs, revitalize local economies, and modernize the energy grid, the group said.
5. Federal government AI adoption strategy
OpenAI said the rate of AI adoption in federal agencies is 'unacceptably low,' pointing out that civil servants, especially those in the national security sector, are not benefiting from AI technology. It called for removing barriers to government AI adoption, such as outdated certification processes, restrictive testing authority, and inflexible procurement paths. Specifically, it proposed modernizing cybersecurity rules for cloud-based applications, accelerating AI testing and experimentation, and implementing rapid procurement mechanisms, and also promoting partnerships to develop custom AI models for national security applications.
By
And in 'Part 2: Export Control Strategy for Democratic AI,' OpenAI states, 'DeepSeek's recent release of the R1 model is notable not for its capabilities, but as an indicator of the state of this competition. While impressive, the R1's inference capabilities are at best on par with several American models.'
By Tim Reckmann
OpenAI warned that 'there are significant risks to building DeepSeek models into critical infrastructure and other high-risk use cases' and pointed out that 'DeepSeek could be coerced by the Chinese Communist Party to manipulate its models to cause harm.'
OpenAI also argues that DeepSeek is 'subsidized, controlled, and free to use by the Chinese Communist Party-led state,' and that the cost users pay is privacy and security. It also claims that DeepSeek must comply with Chinese law to collect user data, and that it uses that data to train more advanced systems for the Chinese Communist Party.
Furthermore, OpenAI argued, 'DeepSeek's models are more willing to generate methods for illegal and harmful activities, such as fraud and intellectual property theft, reflecting the Chinese Communist Party's view of the infringement of American intellectual property rights as a feature rather than a flaw.'
TechCrunch, an IT news site, pointed out that 'it is unclear whether OpenAI's 'model' refers to DeepSeek's API or open model, or both. DeepSeek's open model does not include mechanisms that would allow the Chinese government to extract user data. ' And, 'OpenAI's new allegation that DeepSeek is supported and directed by China intensifies OpenAI's campaign against China.'
Related Posts:
in Software, Posted by log1i_yk