In December 2025, the White House issued an Executive Order establishing a national policy framework for AI. The stated objective is to sustain and enhance US leadership in AI through a minimally burdensome and uniform federal approach.
The Order responds to the growing patchwork of state-level AI regulation and signals a clear federal preference for limiting divergent state experimentation in favour of consistency at the national level. It introduces a framework to identify state AI laws considered unduly burdensome, challenge them through litigation, restrict federal funding to states that maintain such laws and take steps towards federal pre-emption where appropriate. Based on the timelines set out in the Order, implementation is expected to commence within 30 to 90 days.
In November 2025, the White House released a further Executive Order launching the “Genesis Mission”, an initiative aimed at streamlining and accelerating scientific research through artificial intelligence. The Order builds on the administration’s broader AI Action Plan which emphasises the development of high-quality datasets and advanced AI models.
The Genesis Mission seeks to consolidate efforts across national laboratories, research universities and the private sector to develop a collaborative AI platform for scientific discovery. The Order directs the Department of Energy to establish an integrated AI platform drawing on federal scientific datasets to train models capable of testing hypotheses and automating research workflows.
This platform, referred to as the American Science and Security Platform, will leverage national supercomputers, cloud-based computing environments, AI agents, predictive and simulation models, domain-specific models and a mix of proprietary and open-source datasets.
The Order sets out a detailed implementation timetable, requiring the identification of nationally significant scientific challenges within 60 days, available computing resources within 90 days, preliminary datasets and models within 120 days and a demonstration of platform operability within 270 days. The initiative is intended to accelerate discovery, enhance national security and strengthen economic competitiveness.
Recent months have seen the release of high-performing open-source AI models, notably Moonshot AI’s Kimi K2 in November 2025 and DeepSeek V3.2 in December 2025. Both models have demonstrated performance comparable to leading proprietary systems while being distributed under permissive licences.
For organisations considering adoption, deployment architecture remains a key consideration. Self-hosting can enhance data sovereignty and reduce third-party transmission risks but places full compliance and liability responsibility on the deploying organisation. Use via third-party hosting providers may reduce operational burdens but introduces data confidentiality considerations, particularly given longstanding regulatory concerns around data handling by AI service providers.
At federal level, there remains no comprehensive legislation specifically addressing open-source AI. Recent policy initiatives have encouraged the adoption of open-source and open-weight models, highlighting their strategic and innovation value. However, regulatory divergence persists at state level. For example, Colorado’s <span class="news-text_italic-underline">AI Act</span>, due to take effect in February 2026, imposes obligations on developers of high-risk AI systems without an exemption for open-source models. Businesses are therefore advised to assess licensing terms, data governance practices and the evolving compliance landscape carefully.
The United States Patent and Trademark Office has issued revised guidance on inventorship for AI-assisted inventions. The update withdraws earlier policy issued in February 2024 and reaffirms a traditional principle: only natural persons may be named as inventors on US patents and applications.
Under the revised guidance, AI systems are treated as tools used by human inventors. The key question remains whether a human conceived the claimed invention. Where multiple individuals are involved, standard joint inventorship principles apply among the human contributors. No AI-specific test for inventorship has been introduced.
The guidance also clarifies issues around priority and foreign filings. A US application cannot claim priority from an earlier filing that lists only a non-natural person as inventor. Where foreign or international applications list both human and non-human inventors, the US filing should identify only the human inventor or inventors, ensuring continuity of inventorship.
Overall, the updated position confirms that while AI may play a significant role in the inventive process, inventorship under US patent law remains strictly human.
Together, these developments reflect a broader federal strategy focused on regulatory uniformity, accelerated AI-enabled research and clarity on foundational legal concepts such as inventorship. For businesses and researchers operating across borders, the evolving interplay between innovation policy, open-source AI and regulatory compliance will require close and ongoing attention.
