US AI policy

Obligations As a model provider, OpenAI has a responsibility to publicly disclose findings of model misalignment and establish clear standards and timelines for tracking, investigating, and disclosing such issues.

Obligations: Leading AI companies such as Anthropic must grant independent assessors access to internal systems and risk teams and accept their findings.

Obligations: AI companies must follow OpenAI's framework for disclosing model misalignment issues to the public, especially for significant or safety-related findings.

Obligations: Companies must establish internal reporting mechanisms for AI model misalignment, ensuring timely investigation and disclosure. • Opportunities: Early detection and public disclosure of misalignment can enhance model safety and transparency, potentially leading to improved public trust. • Deadlines: Companies need to implement the reporting framework within a reasonable timeframe to comply with OpenAI's new standards.

Healthcare organizations face challenges in AI trust and governance, with 72% using unapproved AI tools.

Obligations: AI developers must ensure models adhere to the principles, cannot be designed to mimic consciousness, and must not be given legal personhood. Models must be within human control and fail tasks that violate the principles.

OpenAI will adjust its R&D strategy to prioritize safety assessments, potentially slowing down the iteration of frontier technologies.

Obligations: Frontier AI developers must comply with consistent safety requirements under a federal framework and undergo independent audits. • Opportunities: Establishing unified safety standards and collaborative frameworks can enhance the safety of AI development and boost industry confidence. • Deadlines: The call for immediate action is made, but specific regulations and frameworks have not yet been issued.

Obligations: Companies developing AI must comply with proposed regulatory guardrails to mitigate risks.

Obligations: AI companies, particularly those in the US, are under pressure to demonstrate responsible development and transparency regarding their AI models.

Companies must master their proprietary tacit knowledge to build continuous learning loops and reduce reliance on single model vendors.

Obligations: AI companies must open their systems to third-party evaluators for safety audits and model alignment assessments, increasing compliance costs and operational complexity.

Obligations: Companies in the AI industry are urged to consider slowing down development and implementing independent monitoring systems. • Opportunities: Early adopters of independent monitoring and transparency measures may gain trust and competitive advantage. • Deadlines: Companies should prepare for potential future regulatory frameworks and industry standards.

Companies must ensure their AI research aligns with the principles of human control and benefit for humanity, avoiding pursuits of meaningless or harmful superintelligence. • Companies should retain control over their unique knowledge and build autonomous continuous learning loops, not becoming dependent on any single model provider. • Microsoft will publish a 'Code of Conduct' for public consultation, providing clear guidance for the industry.

Obligations: Universities must adopt and implement the eight principles to redefine educational goals and assessment methods. Teachers are required to establish clear AI policies for each course and be transparent about their own AI usage.

Obligations: AI data center operators may be subject to new environmental health regulations if the plan is adopted, requiring compliance with stricter air quality standards. • Opportunities: Companies investing in AI data centers in the US could gain from a streamlined regulatory process if the plan is implemented. • Deadlines: The plan's adoption hinges on the president's decision, with no specific timeline mentioned in the report.

Obligations: AI companies may be required to embed auditors and adhere to shared safety standards.

Obligations: AI companies, especially those in the US, must be prepared to respond to official inquiries and provide detailed documentation on AI model security and incident handling procedures. • Opportunities: Companies with robust incident response and documentation practices may be viewed more favorably in regulatory scrutiny. • Deadlines: Companies must ensure timely submission of required documents to avoid potential penalties or reputational damage.

Require policymakers not to treat each innovation in isolation.

Companies may consider the use of copyrighted text for AI training as fair use.

Companies can train AI without excessive concern for copyright infringement risks.

Legalizes the use of copyrighted materials for AI training, providing new opportunities for the AI industry.

Obligations: Developers of frontier AI models must submit their models for a 60-day security assessment 60 days before release, otherwise they may face removal from the market. • Opportunities: Pre-release assessments can reduce security risks for models post-release and enhance public trust. • Deadlines: The policy requires the publication of safety review guidelines by August 31st, but the deadline has passed and the guidelines have not been made public.

Internet-exposed or outdated PLCs are high-risk targets; operators must immediately audit and harden external exposure