Exploring Generative AI’s Threat to Social Justice and Democracy

The talk starts from the idea that personal data is being assembled into "digital twins," detailed profiles of individuals built from purchases, records and everyday interactions. The speakers trace how advertising moved from broadcast messages to demographic and psychographic targeting, then explain at a high level how generative models produce text one word at a time, with biases that are hard to predict or control.

The middle section surveys where these risks already appear. Schools are adopting generative tools faster than rules can follow, hiring platforms screen applicants with opaque models, and healthcare and predictive policing systems reproduce bias found in historical data. The speakers revisit political microtargeting, including the Cambridge Analytica case, and point to deepfake abuse that disproportionately pushes women and minorities out of public life.

The closing part argues that current regulation is late and weak on consent, and that it does not account for the lasting value of data already used to train models. The speakers urge the security community to demonstrate how these systems work, educate friends and family, and engage with policymakers, suggesting that durable protections may need to reach constitutional principles.