AI Pre-Approval: Slows Innovation, No Safety Gain - CDI

Requiring government pre-approval for advanced AI models would slow innovation without making them safer. This is the argument from the Center for Data Innovation. Concerns about AI models, like Claude Mythos, are leading to calls for government review before companies can release new models. These concerns focus on cybersecurity risks, as some AI systems can surpass human ability in finding software vulnerabilities. Some policymakers, like Kevin Hassett from the White House National Economic Council, have suggested requiring AI systems to be "proven safe" before release, similar to FDA review for pharmaceuticals. However, the White House later softened this stance, emphasizing support for innovation over bureaucracy. The Center argues that AI models are information products, not pharmaceuticals, and a pre-approval system would introduce delays, uncertainty, and costs, hindering competition. This approach could also politicize AI development. This matters because it highlights a key debate about how to balance AI safety with technological progress.