Abstract
<p>With the rapid proliferation of AI-generated synthetic mediaacross South Asia, communities in Sri Lanka, India, and Bangladeshface deepfake-related harms that existing Trust &amp; Safety frameworks—designed predominantly for Global North contexts—are ill-equipped toaddress. Through three participatory workshops with 49 stakeholdersincluding journalists, lawyers, fact-checkers, feminist advocates, and civilsociety practitioners, we investigate how deepfake harms are perceived,experienced, and contested across these distinct yet interconnectedcontexts. Our findings reveal four cross-cutting patterns: disproportionategendered targeting of women and marginalized communities; systematicfailure of platform moderation for South Asian languages; inaccessibleinstitutional recourse for victims; and strong community preferencefor locally governed, participatory solutions over externally designedtechnical fixes. We contribute empirical evidence of deepfake harm inunder-researched Global South contexts, a cross-country comparativeframework for Trust &amp; Safety analysis, and a set of design and policyimplications that center community knowledge, survivor support, andlinguistic justice as foundations for a reimagined deepfake governanceparadigm.</p>