Jeffrey S. Collins

The most responsible way to govern powerful technologies like AI is with clear national standards. That’s why we’re pleased that California’s SB 53 created a path toward harmonization with the federal government—the most effective way to pair AI safety with broad access to its benefits. As Gov. Gavin Newsom noted in his signing statement, the federal government plays a meaningful role given AI’s impact on national security and competitiveness. We couldn’t agree more: frontier AI safety is best handled in Washington, while states focus on rules of the road like teen safety or consumer protection. Here’s why: the federal government has access to classified systems and technical capabilities that states do not have. Those allow for models to be tested and potential catastrophic safety issues to be mitigated; states lacking those resources can only hold someone accountable after the harm. It’s in the public’s best interest to focus on tangible harms like fraud, and OpenAI was among the first companies to sign agreements with the US Center for AI Standards and Innovation. CAISI’s work could open the door to national technical standards that balance innovation and safety. In that same spirit, we worked to improve SB 53. The final version lays out a clearer path to harmonize California’s standards with federal ones. That’s also why we support a single federal approach—potentially through the emerging CAISI framework—rather than a patchwork of state laws. We’ve backed up our support for national standards with concrete action on transparency and safety reporting. OpenAI publishes detailed model cards and system cards explaining our technology and outlining its safeguards. We set clear expectations for how our systems should behave through the Model Spec, and we red-team new releases before launch to uncover potential risks. We also publicly share safety evaluations through our Preparedness Framework, which sets out how we measure, monitor, and mitigate large-scale risks. We’ve had these practices in place for years and are glad to see them becoming the industry norm. More broadly, there’s a growing need for a uniform national system. More than 1,000 AI-related bills have been introduced in states this year, creating conflicting rules, friction, and uncertainty. Imagine 50 different states coming up with their own laws to regulate transistors during the Space Race. California will play an outsized role in shaping AI’s future. With its unmatched talent base and tech ecosystem, the right policies can ensure California’s breakthroughs fuel jobs and growth nationwide—and help secure a right to AI that benefits the many, not the few. From railroads to the internet, the US has led by setting clear national guardrails, not letting a fragmented state rules slow progress. AI is no different, and crafting the right policies today will keep it safe, make it easier for developers to innovate, and ensure AI's benefits reach every corner of the country. 

64

3 Comments

New from AidData at William & Mary: Chasing China: Learning to Play by Beijing's Global Lending Rules China’s overseas lending and grant-giving portfolio remains shrouded in secrecy, with questions swirling about its true scale, purpose, and impact. Chasing China sets the record straight with a uniquely comprehensive and granular source of evidence. The authors of the report analyze the full range of projects and activities that Beijing has bankrolled in the developing and developed world since the turn of the century. They do so with AidData’s newly assembled China’s Global Loans and Grants Dataset, Version 1.0, which tracks 30,000 projects and activities across 217 countries financed with grants and loans from 1,193 Chinese official sector donors and lenders worth $2.2 trillion from 2000 to 2023. The report demonstrates that China’s overseas lending portfolio is vastly larger than previously understood. However, its lending for Belt and Road infrastructure projects in the developing world accounts for a surprisingly small share (20%) of the overall portfolio. China’s lending operations in the developed world were previously assumed to be small and insignificant. But the authors find that the share of Beijing’s lending portfolio that supports high-income and upper-middle income countries has skyrocketed from 12% to 76%. Another major discovery is that Chinese state-owned creditors have bankrolled approximately 10,000 projects and activities in 72 high-income countries to the tune of nearly $1 trillion. Beijing has effectively become the new global pace-setter, rewriting the rules and norms that govern the cross-border provision of aid and credit while forcing G7 policymakers to fundamentally rethink their strategy. Its approach to international lending was once a source of scorn, ridicule or bemusement in Western capitals. However, the U.S. and its allies are now seeking to compete with China via emulation rather than differentiation. Rather than continuing to voluntarily tie their hands and limit the provision of aid and credit to high-income countries, G7 countries are loosening the restrictions that prevent their development finance institutions and export credit agencies from supporting projects and activities in high-income countries. At the same time, they are shuttering or slashing the budgets of foreign aid agencies that focus on the needs of low- and middle-income countries. Full report: https://lnkd.in/gzAKDjS2

33

4 Comments