Will global-catastrophic-risk-focused evaluation of certain AI systems by accredited bodies become mandatory in the US before 2035?