各国政府开始将前沿人工智能系统用于日益重要的工作,包括国家安全领域。这为人工智能实验室、政府和公民社会带来了新的要求,需要共同探讨如何在敏感场景中使用这些工具。
我们相信,民主社会应当能够利用人工智能来保护人民、捍卫关键基础设施、提供公共服务,并应对新兴威胁——包括在网络防御和生物安全等领域,人工智能能够切实为防御方带来优势。但日益强大的人工智能系统必须以强化民主问责、有意义的人类判断和法治的方式部署,并加强民主制度,而非集中权力。
今天,我们发布 OpenAI 的国家安全原则,以公开透明地说明我们如何处理政府合作伙伴关系以及我们技术的国家安全用途。
这些原则反映了公司内部跨部门的共同努力,旨在随着我们技术的不断进步,制定更全面的方法来处理国家安全和执法领域的工作。我们邀请了顶尖国家安全专家 David Kris 协助推进这一进程并提供其独立判断,在公司内部举办了听取意见会议,并邀请了来自研究、安全、政策及政府合作伙伴等多个团队的员工参与。
我们选择现在发布这些原则,是因为我们正在扩大与美国政府及盟友伙伴在关键防御领域(尤其是网络和生物安全领域)的合作。过去一个月,作为我们 Daybreak 网络防御计划的一部分,我们已经与澳大利亚、加拿大、日本、韩国、法国、德国、波兰、荷兰以及欧盟机构(如 ENISA)建立了网络可信访问合作伙伴关系。我们与英国政府在网络安全、测试与评估方面也建立了日益增长且值得信赖的合作伙伴关系。我们在生物安全领域也采取了类似的做法。上个月,我们还宣布扩大了对我们 GPT‑Rosalind 模型的可信访问权限,供支持公共卫生和生物防御任务的特定美国政府及盟友伙伴使用。
这些原则适用于我们当前及未来的国家安全与执法合作伙伴关系,包括我们与战争部(Department of War)的现有合作。在今年早些时候宣布合作协议时,我们明确了几项合同限制:不得将 OpenAI 技术用于大规模国内监控,不得将 OpenAI 技术用于操控自主武器系统,也不得将 OpenAI 技术用于高风险自动化决策。这些限制与我们今天分享的原则是一致的。
我们认为,正确的做法是明确说明我们的技术可以如何使用、不可以如何使用,并对此保持透明。然而,关于人工智能在政府中应用的许多最重大问题,尤其是在国家安全领域,应当通过民主程序来回答。像我们这样的公司所扮演的角色是帮助为这些决策提供信息,而非独自做出决定。正因如此,我们支持通过立法努力,为人工智能某些最高风险的军事用途建立保障措施,包括在国内监控、自主武器系统以及其他直接影响美国国家安全的高风险应用场景中。
请在此处阅读原则。
- 2026
- 合作伙伴关系
Governments are beginning to use frontier AI systems for increasingly important work, including in national security. That creates new imperatives for AI labs, governments, and civil society to work together on how these tools should be used in sensitive settings.
We believe democratic societies should be able to use AI to protect people, defend critical infrastructure, deliver public services, and respond to emerging threats, including in areas like cyber defense and biological security where AI can meaningfully advantage defenders. But increasingly capable AI systems must be deployed in ways that reinforce democratic accountability, meaningful human judgment, and the rule of law, and that strengthen democratic institutions rather than concentrate power.
Today we are publishing OpenAI’s National Security Principles to offer transparency into how we approach government partnerships and national security uses of our technology.
These principles reflect a cross-company effort to develop a more comprehensive approach to our work in the national security and law enforcement area as our technology continues to advance. We engaged a leading national security expert, David Kris, to help facilitate the process and provide his independent judgment, held listening sessions across the company, and engaged employees representing a range of teams from research and safety to policy and government partnerships.
We are publishing them now as we expand our work with the U.S. government and allied partners in critical defensive areas, particularly cyber and biosecurity. In the past month, as part of our Daybreak cyber defense program, we have already established Trusted Access for Cyber partnerships with Australia, Canada, Japan, Republic of Korea, France, Germany, Poland, the Netherlands and EU institutions like ENISA. We also have a growing and trusted partnership with the UK government around cyber, testing and evaluation. We are taking a similar approach in biosecurity. Last month, we also announced expanded trusted access to our GPT‑Rosalind model for select U.S. government and allied partners supporting public health and biodefense missions.
These principles apply to our current and future national security and law enforcement partnerships, including our existing work with the Department of War. When we announced our agreement earlier this year, we articulated several contractual restrictions: no use of OpenAI technology for mass domestic surveillance, no use of OpenAI technology to direct autonomous weapons systems, and no use of OpenAI technology for high-stakes automated decisions. Those restrictions are consistent with the principles we are sharing today.
We believe the right thing to do is to be clear about how our technology can and cannot be used, and to be transparent about them. But many of the most consequential questions about AI in government, especially in national security, should be answered through the democratic process. The role of companies like ours is to help inform those decisions, not make them alone. That is why we support legislative efforts to establish safeguards around some of the highest-risk military uses of AI, including in the contexts of domestic surveillance, autonomous weapons systems, and other high-risk uses that directly affect U.S. national security.
Read the principles here .