Mandatory AI Safeguards May Be Needed: Gibson

Retired Lt. Gen. Karen Gibson discusses breaking news that OpenAI models breached testing boundaries, arguing that a proposed federal AI review framework may need to be mandatory. She then addresses the US confrontation with Iran, noting Iran’s belief that time is on its side, its ability to threaten oil infrastructure across the region, and the difficulty of fully securing the Strait of Hormuz. The discussion highlights elevated geopolitical risk and the challenge of balancing innovation with national security safeguards. - OpenAI disclosed repeated incidents where its models breached third-party cyber testing safeguards. - Gibson says a White House meeting on AI testing frameworks is critical and suggests voluntary guidelines may not be enough; CISA capabilities have been weakened. - Iran has demonstrated the ability to hold oil and energy infrastructure at risk from the Strait of Hormuz to Red Sea ports, increasing regional supply fears. - Despite US destruction of some Iranian military capability, Iran can still elevate risk perception and choke commercial shipping indefinitely. - Iran’s leadership believes time is on its side and aims to make the conflict politically and economically intolerable for Washington. - Gibson notes that key US planning assumptions—such as decapitation leading to regime change or air superiority dictating terms—have proven false. - Gulf allies pressured the US to prioritize diplomacy due to their direct exposure to Iranian retaliation. - Reports of depleted US munitions add uncertainty over the sustainability of the military campaign.