美国 AI 政策

AI公司必须确保其模型不会发动或协助网络攻击，否则可能面临法律追责

公司若需获取Reddit数据必须与Reddit签署商业合作协议

AI公司必须配合调查并提供内部记录和高管证词

AI实验室需配合调查并提供相关文件和高管证词

为公民提供更便捷的政府服务入口，简化跨机构信息获取流程

智驾系统供应商需加强针对静止或低速车辆的探测能力

AI公司需评估其产品限制是否符合供应链安全法规要求，避免被认定为'操纵产品'

义务：美国航空局（FAA）必须确保其部署的AI系统（如SMART）在机场航线管理中符合安全标准，并向空中交通管制员提供可靠的建议路线，但系统本身不直接控制飞机。• 机会：AI系统可帮助优化空中交通流量，减少拥堵，提升机场运营效率。• 截止日期：系统于2025年1月导致67人死亡的空难发生地投入使用，需确保其安全性和有效性。

Obligations: Companies deploying Meta smart glasses must comply with Meta's privacy protection guidelines, ensuring proper implementation of the indicator light system and promptly applying software updates to prevent vulnerabilities.

Obligations: Companies must comply with the new AI-based air traffic management system for flight scheduling and rerouting. • Opportunities: Enhanced flight safety, increased air traffic capacity, and improved operational efficiency. • Deadlines: The system is set to be operational by September 21, 2026, with a 12-year contract for development.

必须为AI模型配备‘kill switch’功能并设立独立审计员； - 需在两个月内配合专家小组提供相关建议； - 需遵守加州AI安全监管要求，否则面临合规风险。

义务：OpenAI 必须按照新框架的要求，追踪、调查并公开其模型对齐问题，设定了明确的披露标准和时间线。

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.

公司义务：Cognita Imaging需在18个月内完成1百万份放射报告的LLM评估方法开发与验证工作，并符合FDA的研究要求。

Obligations: Companies deploying macOS systems must promptly apply the security updates to mitigate risks of kernel memory corruption and privilege escalation. • Opportunities: Engaging with international security research teams, such as those from China, can enhance the discovery and reporting of vulnerabilities. • Deadlines: Immediate action is required to apply the updates to protect against known vulnerabilities.

律师在使用AI生成法律文书时，必须严格核实内容的真实性，避免提交虚假信息，否则可能面临藐视法庭的处罚和罚款。

ARPA-H投资AI治疗心力衰竭项目；

对公司：提供了一种在获得正式营销授权前向市场推出产品的途径。

对公司而言，使用受版权保护的文本训练AI模型可能被视为合理使用。

公司在进行AI训练时无需过度担忧版权侵权风险。

对公司而言，使用受版权保护的材料进行AI训练合法化。