EthosGPT: Whose Values Guide Technological Change? Cultural Representation, Language-Model Updates, and Creative Destruction

Technological innovation can raise product quality while replacing established work, the creative-destruction process honored by the 2025 economics Nobel. If investors and support agencies consult AI advisers, whose views guide innovation and help for those affected? We compare archived GPT-5.5 and GPT-5.6 Sol answers to six World Values Survey questions with unweighted survey summaries from 64 countries. The newer model lowers average category-share error by 0.0080 (95% country-bootstrap interval [-0.0107, -0.0050]), yet its directional gaps vary across questions and places. An illustrative adviser rule translates these gaps into expected product-quality gain (an efficiency channel) and replacement without assumed transition help (an equity-risk channel). Nigeria's quality shortfall widens and assumed help weakens; Kenya's quality advantage coexists with greater unassisted exposure. These scenarios are not observed growth or employment effects. The paired audit and explicit decision rule show why a better average survey match could still lead to different choices about who gains and who bears adjustment costs. Locally validated questions, observed decisions, and longitudinal livelihood and environmental data could test when AI-assisted innovation contributes to sustainable growth.