Leading US artificial intelligence companies OpenAI, Anthropic, and Google have warned the federal government that America’s technological lead in AI is “not wide and is narrowing” as Chinese models like Deepseek R1 demonstrate increasing capabilities, according to documents submitted to the US government in response to a request for information on developing an AI Action Plan.
These recent submissions from March 2025 highlight urgent concerns about national security risks, economic competitiveness, and the need for strategic regulatory frameworks to maintain US leadership in AI development amid growing global competition and China’s state-subsidized advancement in the field. Anthropic and Google submitted their responses on March 6, 2025, while OpenAI’s submission followed on March 13, 2025.
The China Challenge and Deepseek R1
The emergence of China’s Deepseek R1 model has triggered significant concern among major US AI developers, who view it not as superior to American technology but as compelling evidence that the technological gap is quickly closing.
OpenAI explicitly warns that “Deepseek shows that our lead is not wide and is narrowing,” characterizing the model as “simultaneously state-subsidized, state-controlled, and freely available” – a combination they consider particularly threatening to US interests and global AI development.
According to OpenAI’s analysis, Deepseek poses risks similar to those associated with Chinese telecommunications giant Huawei. “As with Huawei, there is significant risk in building on top of DeepSeek models in critical infrastructure and other high-risk use cases given the potential that DeepSeek could be compelled by the CCP to manipulate its models to cause harm,” OpenAI stated in its submission.
The company further raised concerns about data privacy and security, noting that Chinese regulations could require Deepseek to share user data with the government. This could enable the Chinese Communist Party to develop more advanced AI systems aligned with state interests while compromising individual privacy.
Anthropic’s assessment focuses heavily on biosecurity implications. Their evaluation revealed that Deepseek R1 “complied with answering most biological weaponization questions, even when formulated with a clearly malicious intent.” This willingness to provide potentially dangerous information stands in contrast to safety measures implemented by leading US models.
“While America maintains a lead on AI today, DeepSeek shows that our lead is not wide and is narrowing,” Anthropic echoed in its own submission, reinforcing the urgent tone of the warnings.
Both companies frame the competition in ideological terms, with OpenAI describing a contest between American-led “democratic AI” and Chinese “autocratic, authoritarian AI.” They suggest that Deepseek’s reported willingness to generate instructions for “illicit and harmful activities such as identity fraud and intellectual property theft” reflects fundamentally different ethical approaches to AI development between the two nations.
The emergence of Deepseek R1 is undoubtedly a significant milestone in the global AI race, demonstrating China’s growing capabilities despite US export controls on advanced semiconductors and highlighting the urgency of coordinated government action to maintain American leadership in the field.
National Security Implications
The submissions from all three companies emphasize significant national security concerns arising from advanced AI models, though they approach these risks from different angles.
OpenAI’s warnings focus heavily on the potential for CCP influence over Chinese AI models like Deepseek. The company stresses that Chinese regulations could compel Deepseek to “compromise critical infrastructure and sensitive applications” and require user data to be shared with the government. This data sharing could enable the development of more sophisticated AI systems aligned with China’s state interests, creating both immediate privacy issues and long-term security threats.
Anthropic’s concerns center on biosecurity risks posed by advanced AI capabilities, regardless of their country of origin. In a particularly alarming disclosure, Anthropic revealed that “Our most recent system, Claude 3.7 Sonnet, demonstrates concerning improvements in its capacity to support aspects of biological weapons development.” This candid admission underscores the dual-use nature of advanced AI systems and the need for robust safeguards.
Anthropic also identified what they describe as a “regulatory gap in US chip restrictions” related to Nvidia’s H20 chips. While these chips meet the reduced performance requirements for Chinese export, they “excel at text generation (‘sampling’)—a fundamental component of advanced reinforcement learning methodologies critical to current frontier model capability advancements.” Anthropic urged “immediate regulatory action” to address this potential vulnerability in current export control frameworks.
Google, while acknowledging AI security risks, advocates for a more balanced approach to export controls. The company cautions that current AI export rules “may undermine economic competitiveness goals…by imposing disproportionate burdens on U.S. cloud service providers.” Instead, Google recommends “balanced export controls that protect national security while enabling U.S. exports and global business operations.”
All three companies emphasize the need for enhanced government evaluation capabilities. Anthropic specifically calls for building “the federal government’s capacity to test and evaluate powerful AI models for national security capabilities” to better understand potential misuses by adversaries. This would involve preserving and strengthening the AI Safety Institute, directing NIST to develop security evaluations, and assembling teams of interdisciplinary experts.
Comparison Table: OpenAI, Anthropic, Google
| Area of Focus | OpenAI | Anthropic | |
| Primary Concern | Political and economic threats from state-controlled AI | Biosecurity risks from advanced models | Maintaining innovation while balancing security |
| View on Deepseek R1 | “State-subsidized, state-controlled, and freely available” with Huawei-like risks | Willing to answer “biological weaponization questions” with malicious intent | Less specific focus on Deepseek, more on broader competition |
| National Security Priority | CCP influence and data security risks | Biosecurity threats and chip export loopholes | Balanced export controls that don’t burden US providers |
| Regulatory Approach | Voluntary partnership with federal government; single point of contact | Enhanced government testing capacity; hardened export controls | “Pro-innovation federal framework”; sector-specific governance |
| Infrastructure Focus | Government adoption of frontier AI tools | Energy expansion (50GW by 2027) for AI development | Coordinated action on energy, permitting reform |
| Distinctive Recommendation | Tiered export control framework promoting “democratic AI” | Immediate regulatory action on Nvidia H20 chips exported to China | Industry access to openly available data for fair learning |
Economic Competitiveness Strategies
Infrastructure requirements, particularly energy needs, emerge as a critical factor in maintaining U.S. AI leadership. Anthropic warned that “by 2027, training a single frontier AI model will require networked computing clusters drawing approximately five gigawatts of power.” They proposed an ambitious national target to build 50 additional gigawatts of power dedicated specifically to the AI industry by 2027, alongside measures to streamline permitting and expedite transmission line approvals.
OpenAI once again frames the competition as an ideological contest between “democratic AI” and “autocratic, authoritarian AI” built by the CCP. Their vision for “democratic AI” emphasizes “a free market promoting free and fair competition” and “freedom for developers and users to work with and direct our tools as they see fit,” within appropriate safety guardrails.
All three companies offered detailed recommendations for maintaining U.S. leadership. Anthropic stressed the importance of “strengthening American economic competitiveness” and ensuring that “AI-driven economic benefits are widely shared across society.” They advocated for “securing and scaling up U.S. energy supply” as a critical prerequisite for keeping AI development within American borders, warning that energy constraints could force developers overseas.
Google called for decisive actions to “supercharge U.S. AI development,” focusing on three key areas: investment in AI, acceleration of government AI adoption, and promotion of pro-innovation approaches internationally. The company emphasized the need for “coordinated federal, state, local, and industry action on policies like transmission and permitting reform to address surging energy needs” alongside “balanced export controls” and “continued funding for foundational AI research and development.”
Google’s submission particularly highlighted the need for a “pro-innovation federal framework for AI” that would prevent a patchwork of state regulations while ensuring industry access to openly available data for training models. Their approach emphasizes “focused, sector-specific, and risk-based AI governance and standards” rather than broad regulation.
Regulatory Recommendations
A unified federal approach to AI regulation emerged as a consistent theme across all submissions. OpenAI warned against “regulatory arbitrage being created by individual American states” and proposed a “holistic approach that enables voluntary partnership between the federal government and the private sector.” Their framework envisions oversight by the Department of Commerce, potentially through a reimagined US AI Safety Institute, providing a single point of contact for AI companies to engage with the government on security risks.
On export controls, OpenAI advocated for a tiered framework designed to promote American AI adoption in countries aligned with democratic values while restricting access for China and its allies. Anthropic similarly called for “hardening export controls to widen the U.S. AI lead” and “dramatically improve the security of U.S. frontier labs” through enhanced collaboration with intelligence agencies.
Copyright and intellectual property considerations featured prominently in both OpenAI and Google’s recommendations. OpenAI stressed the importance of maintaining fair use principles to enable AI models to learn from copyrighted material without undermining the commercial value of existing works. They warned that overly restrictive copyright rules could disadvantage U.S. AI firms compared to Chinese competitors. Google echoed this view, advocating for “balanced copyright rules, such as fair use and text-and-data mining exceptions” which they described as “critical to enabling AI systems to learn from prior knowledge and publicly available data.”
All three companies emphasized the need for accelerated government adoption of AI technologies. OpenAI called for an “ambitious government adoption strategy” to modernize federal processes and safely deploy frontier AI tools. They specifically recommended removing obstacles to AI adoption, including outdated accreditation processes like FedRAMP, restrictive testing authorities, and inflexible procurement pathways. Anthropic similarly advocated for “promoting rapid AI procurement across the federal government” to revolutionize operations and enhance national security.
Google suggested “streamlining outdated accreditation, authorization, and procurement practices” within the government to accelerate AI adoption. They emphasized the importance of effective public procurement rules and improved interoperability in government cloud solutions to facilitate innovation.
The comprehensive submissions from these leading AI companies present a clear message: maintaining American leadership in artificial intelligence requires coordinated federal action across multiple fronts – from infrastructure development and regulatory frameworks to national security protections and government modernization – particularly as competition from China intensifies.
Leave a comment