Independent AI audits make sense, but should not start too early: that is SIIA’s position in a statement published on October 7. Paul Lekas, its executive vice president for global public policy and government affairs, was responding to Senator Maria Cantwell’s proposed oversight framework for frontier AI.
Lekas supports the audit approach but argues that the necessary benchmarks and standards are not yet in place. He calls for more attention and investment in measurement science at NIST and other organizations. This is the association’s assessment, not an independent inventory of all existing AI standards.
Cantwell proposes that covered models pass an independent audit against safety standards before release. Those standards would address risks including loss of control and agents leaving secure testing environments.
Her framework remains a proposal; Lekas’ response contains neither an investment commitment nor a shared industry agreement.