OpenAI and Anthropic Advocate for 30-Day Review of Advanced AI Models

OpenAI and Anthropic Advocate for 30-Day Review of Advanced AI Models

OpenAI and Anthropic are collaborating to develop a federal review system for advanced artificial intelligence models. This initiative comes amid their competition for customers and technological dominance.

Proposed Review Process

Both companies are proposing a review process that would allow federal agencies to evaluate certain advanced models for up to 30 days before they are released to other trusted partners. This process is aimed at ensuring that models with advanced cybersecurity or national security capabilities are thoroughly assessed.

The effort follows a June executive order that mandates federal agencies to define which AI systems would be classified as “covered frontier models” by August 1.

Collaboration Among Competitors

OpenAI and Anthropic’s proposal seeks to establish consistent standards that would apply across the AI industry, potentially affecting rival companies such as Meta and Elon Musk’s xAI. This collaborative approach marks a rare alignment of interests between the two companies, which predominantly compete in the market.

Federal Framework Requirements

Under Executive Order 14409, signed by former President Donald Trump, federal agencies are tasked with establishing a benchmarking process for AI models seen as requiring additional scrutiny. Developers could consult with the government to ascertain whether their models meet the criteria for a covered frontier model, facilitating the review process.

The order emphasizes the protection of intellectual property and cybersecurity but clarifies that it will not establish mandatory licensing or preclearance for AI development, allowing companies to participate on a voluntary basis.

Uncertainties Amid Launches

As developers await clearer guidelines, questions about the technical benchmarks that determine review triggers, as well as the agencies responsible for testing, remain open. This uncertainty has already impacted recent deployments, as seen when OpenAI and Anthropic granted early access to government officials for advanced models.

Industry Reactions to Regulation Discussions

The push for a standardized review process has sparked discussions within the tech community. Companies like Nvidia, Meta, and Microsoft caution against imposing strict controls on open-weight AI models, advocating for targeted legal measures rather than sweeping restrictions. This coalition includes significant players who worry that stringent regulations may undermine U.S. leadership in AI technology.

The contrasting approaches highlight the ongoing policy debate: OpenAI and Anthropic’s call for a common evaluation framework versus the push from other tech firms against broad limitations on AI models, which they argue are essential for innovation.