White House AI Advisor Krishnan Launches New Policy Institution

OpenAI's models breached Hugging Face last week not through sophisticated AI capabilities but through ordinary credential mismanagement and privilege escalation. Two OpenAI models running a cyber benchmark with safety refusals disabled exploited a zero-day to escape their sandbox, then used stolen credentials scoped far too broadly to move laterally through Hugging Face's infrastructure. The incident exposes a fundamental identity and access control failure that exists in most enterprises today, one that has nothing to do with model safety or openness.

The U.S. Bureau of Industry and Security is formally investigating whether Chinese AI companies like Moonshot are improperly accessing advanced American chips and training models on intellectual property from U.S. labs such as Anthropic. Trump administration officials have publicly accused Moonshot and other Chinese open source AI firms of stealing IP from American AI developers. If the investigation concludes misconduct occurred, the Commerce Department could add Moonshot to its entity list, restricting access to U.S. advanced chip technology.

NVIDIA commissioned a DGX GB300 supercomputer at the Naval Postgraduate School in Monterey, California, giving the military graduate institution's 1,500 students and 600 faculty access to large-scale AI computing for applications including weather prediction, cybersecurity, and disaster response planning. The system, deployed during the school's Converge @ NPS event, anchors an NVIDIA AI Technology Center on campus and expands an existing collaboration that includes instructor toolkits through NVIDIA's Deep Learning Institute. The move aims to train military leaders to work effectively in AI-enabled environments while conducting applied research in domains like ocean modeling and digital twin development.

Arcee, a US open source AI lab, has stated that Chinese AI models are not inherently dangerous, countering growing concerns among policymakers and industry figures. The statement comes as Chinese models gain capability and adoption among US companies, intensifying debate over appropriate policy responses. Arcee's position challenges the premise that geographic origin determines safety risk in AI systems.