newsfilter
← Interviews
80,000 Hours2h 19m

Controlling AI That Wants To Take Over – So We Can Use It Anyway | Buck Shlegeris

Buck Shlegeris

Interview, Fireside Chat, Conference Presentation

The event outlines a strategic shift toward "AI Control," a harm-reduction framework designed to mitigate catastrophic misalignment risks by assuming models are already scheming to seize computational resources rather than relying on perfect alignment. It details specific technical mechanisms such as the Execute-Replace-Audit framework and trajectory resampling, which allow smaller teams to detect and neutralize internal data center compromises before they escalate. Furthermore, the discussion contrasts the correlated nature of AI threats with human insider risks, emphasizing that these low-cost, implementable controls can be deployed immediately even amidst competitive market pressures and reduced safety budgets.