TRUMP'S AI TESTING PLAN EXCLUDES OPEN MODELS

The voluntary guidelines, created following a Trump executive order, contain significant gaps in scope. The framework refuses to test open models—software anyone can download and inspect—and explicitly states it cannot be used to restrict open models once released. This approach raises questions about comprehensive AI risk assessment. Open models represent a growing segment of AI development, with major releases from Meta, Mistral, and others making powerful systems widely available. The administration's framework focuses on proprietary models instead, where testing and oversight may be more feasible. However, excluding open models leaves potential cybersecurity vulnerabilities unexamined across a significant portion of the AI landscape. The voluntary nature of the guidelines further limits their scope, as participation depends on company cooperation rather than mandatory compliance. Security experts have not yet assessed whether this approach adequately addresses risks posed by advanced AI systems across all deployment types.