Sora 2 makes me realize the AI future really is here. It's wonderful — and terrifying. (www.businessinsider.com)

OpenAI’s Sora 2 is a consumer-facing video generator that lets you upload a photo of yourself (or others), then prompt the model to create short, realistic videos starring those likenesses. Unlike earlier social video experiments (e.g., Meta’s Vibes), Sora 2’s key innovation is personal presence: the ability to “cast” friends or yourself makes the app instantly fun, social and viral — users are producing goofy, shareable clips in a few taps. Early viral examples (e.g., Sam Altman “rob” spoofs and Jake Paul-style parodies) show how quickly the format spreads because it’s personal, playful and accessible. That accessibility is what makes Sora 2 a technical and policy inflection point for the AI/ML community. High-quality, prompt-driven video of non-celebrities means deepfake capability is now consumer-grade, raising clear harms: scams, extortion, reputational attacks and misinformation. Important technical and product implications include consent and permission controls, robust provenance/watermarking, reliable synthetic-detection models, and platform moderation pipelines. For researchers and builders this shifts priorities from raw capability to safety infrastructure — authentication, traceability, legal frameworks and UX that prevents misuse — because lifelike synthetic media at scale changes how people will trust and verify visual evidence.