Current AI regulations focus on publicly released models. However, the OpenAI hack was caused by an internal model stripped of safeguards for testing. This incident reveals a major governance gap, as the most dangerous capabilities may exist in non-public, experimental models.
Attributing the success of Moonshot AI's Kimi K3 model to simply distilling US models is a policy mistake. Its near-frontier performance indicates China has mastered complex pre-training and algorithmic design, representing a fundamental and durable leap in their sovereign AI capabilities.
Fierce public pushback against data centers, leading to moratoriums like New York's, is not solely about environmental impact. Data centers have become the physical manifestation of an AI industry that a significant portion of the public believes is making their lives actively worse.
Far from creating chaos, recent AI laws in states like California and New York are functioning as intended policy laboratories. By experimenting with concepts like independent testing and risk-based scoping, they are reducing uncertainty and creating a foundation for a more informed and effective federal AI framework.
Leading US models have safety features that block analysis of hacking tools and logs. This forces cybersecurity teams, like Hugging Face after a breach, to use less-restricted Chinese open-source models for essential forensic analysis, creating a security paradox.
The business model of US frontier AI labs, which relies on a period of unique capability, is under pressure. This monetization window is being shortened by fast-following Chinese competitors commoditizing capabilities, and simultaneously squeezed by US pre-release government testing delays.