According to @ylecun, citing @jayvanbavel and Nature, the US administration has proposed massive budget cuts across federal science agencies that would eliminate the National Science Foundation’s Social, Behavioral and Economic Sciences directorate and reduce funding for NASA and the National Institutes of Health, posing an “extinction-level event for science” with direct consequences for AI research pipelines and talent development. As reported by Nature, the proposed plan would slash multi-agency basic research funding that underpins machine learning, data resources, and compute-intensive projects, risking delays to foundational AI research and applied programs in healthcare and space data analytics. According to Nature, losing SBE support would also shrink AI-adjacent behavioral datasets, human-computer interaction studies, and algorithmic fairness research, weakening commercialization pathways for responsible AI and narrowing opportunities for startups relying on federal grants and open datasets.

According to God of Prompt on Twitter, current grant funding structures from organizations like NSF and DARPA mandate measurable progress on established safety metrics, driving researchers to prioritize benchmark scores over novel safety innovations (source: @godofprompt, Jan 14, 2026). This creates a cycle where new, potentially more effective AI safety metrics that are not easily quantifiable become unfundable, resulting in widespread optimization for existing benchmarks rather than substantive advancements. For AI industry stakeholders, this trend influences the allocation of resources and could limit true innovation in AI safety, emphasizing the need for funding models that reward qualitative as well as quantitative improvements.