OpenAI's AI Regulation Blueprint: A New Proposal

OpenAI Presents an AI Economic Strategy

On Monday, OpenAI released what it terms an “economic blueprint” for Artificial Intelligence. This document, intended as a continually updated resource, outlines policy recommendations for collaboration between the company, the U.S. government, and allied nations.

The Need for Investment in AI

The blueprint, featuring a foreword by Chris Lehane, OpenAI’s VP of global affairs, emphasizes the necessity for the U.S. to attract substantial investment. This funding is crucial for securing the chips, data, energy, and talent required to achieve leadership in the field of AI.

Lehane asserts that the U.S. government has an opportunity to foster the growth of its AI industry. This would simultaneously bolster innovation and safeguard national security, particularly as other nations may be hesitant to fully embrace AI’s potential.

Calls for Government Action and Infrastructure

OpenAI has consistently urged the U.S. government to implement more decisive measures regarding AI. Specifically, the company advocates for increased infrastructure support to facilitate the technology’s ongoing development.

Currently, AI regulation is largely delegated to individual states. OpenAI views this fragmented approach as unsustainable, as evidenced by the proliferation of conflicting state-level bills.

Concerns Regarding Existing Legislation

In 2024 alone, nearly 700 AI-related bills were introduced by state lawmakers, creating a complex and often contradictory legal landscape. For instance, Texas’ Responsible AI Governance Act introduces significant liability concerns for developers of open-source AI models.

OpenAI CEO Sam Altman has also expressed criticism of federal laws like the CHIPS Act. He believes the Act has not been as successful as anticipated in revitalizing the U.S. semiconductor industry and suggests potential improvements under a future administration.

Challenges in U.S. Infrastructure Development

Altman highlighted the difficulties of building infrastructure within the United States. He noted that constructing facilities like power plants and data centers has become excessively bureaucratic, hindering the nation’s ability to lead in AI development.

Power and Data Infrastructure Requirements

To support the data centers essential for AI development and operation, OpenAI’s blueprint recommends a significant increase in federal funding. This funding should be directed towards both power infrastructure and data transmission capabilities.

The blueprint also calls for the development of “new energy sources,” including solar, wind, and nuclear power. OpenAI, alongside other AI companies, has previously advocated for nuclear energy to meet the substantial electricity demands of advanced server farms.

Policy Recommendations for Responsible AI Deployment

In the short term, OpenAI proposes that the government establish “best practices” for AI model deployment. This would help mitigate potential misuse and streamline engagement between the AI industry and national security agencies.

Furthermore, the blueprint suggests developing export controls that facilitate model sharing with allies while restricting access for “adversary nations.” It also encourages information sharing between the government and AI vendors regarding national security threats and resource allocation for risk assessment.

Streamlining Safety and Security Measures

The blueprint emphasizes the need for a streamlined federal approach to frontier model safety and security. It advocates for responsible export of AI models to allies, fostering their own AI ecosystems and reliance on U.S. technology.

Existing Partnerships and Future Collaboration

OpenAI already collaborates with several U.S. government departments. Should its blueprint gain traction, the company anticipates expanding these partnerships.

Current collaborations include cybersecurity work with the Pentagon and a partnership with defense startup Anduril to provide AI technology for countering drone attacks.

International Standards and Voluntary Pathways

OpenAI calls for the development of standards “recognized and respected” internationally, representing the interests of the U.S. private sector. However, the company stops short of advocating for mandatory regulations.

Instead, the blueprint proposes a “defined, voluntary pathway” for companies to collaborate with the government. This collaboration would involve defining model evaluations, testing, and information exchange to enhance safeguards.

Alignment with the Biden Administration and Potential Shifts

The Biden administration’s AI executive order adopted a similar approach, focusing on voluntary safety and security standards. This order established the U.S. AI Safety Institute (AISI), which partners with companies like OpenAI to evaluate model safety.

However, the potential for a change in administration raises concerns, as Trump and his allies have indicated their intention to repeal Biden’s executive order, potentially jeopardizing the AISI.

Addressing Copyright Concerns in AI Development

OpenAI’s blueprint also addresses the complex issue of copyright as it relates to AI. The company argues that AI developers should have access to “publicly available information,” including copyrighted content, for model development.

OpenAI trains its models on publicly available data, including copyrighted material, and has established licensing agreements and opt-out options for creators. However, the company acknowledges the necessity of using copyrighted materials for effective AI training, leading to legal challenges from creators.

Maintaining a Competitive Edge in AI

The blueprint highlights the risk of other nations disregarding intellectual property rights. It argues that if the U.S. and like-minded countries don’t address this imbalance, content will still be used for AI training elsewhere, benefiting other economies.

The government should ensure AI can learn from universal information, similar to humans, while protecting creators from unauthorized digital replicas.

OpenAI’s Increased Lobbying and Strategic Hires

OpenAI has significantly increased its lobbying efforts, spending $800,000 in the first half of last year compared to $260,000 in all of 2023. The company has also recruited former government leaders to its executive team.

These hires include Sasha Baker (ex-Defense Department), Paul Nakasone (former NSA chief), and Aaron Chatterji (formerly the chief economist at the Commerce Department).

Advocacy for Specific Legislation

OpenAI has become more vocal about its preferred AI laws and rules, supporting Senate bills that would establish a federal rule-making body for AI and provide funding for AI research and development.

The company has also opposed legislation it believes would stifle innovation, such as California’s SB 1047.

Stay Informed with TechCrunch’s AI Newsletter

For further insights into the world of AI, subscribe to TechCrunch’s AI-focused newsletter, delivered every Wednesday.