Saved in:
Bibliografiske detaljer
Main Authors: Real Safety AI Foundation, Gilly, Travis
Format: Recurso digital
Sprog:
Udgivet: Zenodo 2026
Fag:
Online adgang:https://doi.org/10.5281/zenodo.19242567
Tags: Tilføj Tag
Ingen Tags, Vær først til at tagge denne postø!
_version_ 1866902083554246656
author Real Safety AI Foundation
Gilly, Travis
author_facet Real Safety AI Foundation
Gilly, Travis
contents <p><strong>Abstract</strong></p> <p>In January 2026, we published a convergence threat analysis demonstrating that ambient<br>non-consensual intimate image synthesis via AR wearables was architecturally feasible through<br>cloud-assisted pipelines and would achieve edge-only feasibility within 12 to 24 months [1]. This<br>technical update reports that multiple independent developments in the eight weeks following<br>publication have compressed our timeline estimates significantly. Specifically, we identify three<br>capabilities released between February and March 2026 that collectively reduce barriers across the<br>critical pipeline stages: person segmentation (MatAnyone2, 140MB model achieving state-of-the-art<br>video matting), inference acceleration (Diagonal Distillation, achieving 270x speedup over baseline<br>video generation), and mobile 3D rendering (MobileGS, achieving 120+ FPS Gaussian splatting on a<br>Snapdragon 8 Gen 3 phone at 4.8MB model size). We revise our feasibility classes accordingly, noting<br>that the "edge full video synthesis" class we originally projected at 12 months may now be achievable<br>within 6 to 9 months under moderate assumptions. We discuss implications for the policy<br>recommendations in our original analysis and argue that the structural mismatch between capability<br>maturation and regulatory response has widened.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_19242567
institution Zenodo
language
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle Compressed Feasibility: A Technical Update to the Ambient Non-Consensual Synthesis Threat Model
Real Safety AI Foundation
Gilly, Travis
Threat analysis
AI safety
deepfakes
perceptual consent
image-based sexual abuse
adversarial machine learning
augmented reality
AR wearables
Threat modeling
<p><strong>Abstract</strong></p> <p>In January 2026, we published a convergence threat analysis demonstrating that ambient<br>non-consensual intimate image synthesis via AR wearables was architecturally feasible through<br>cloud-assisted pipelines and would achieve edge-only feasibility within 12 to 24 months [1]. This<br>technical update reports that multiple independent developments in the eight weeks following<br>publication have compressed our timeline estimates significantly. Specifically, we identify three<br>capabilities released between February and March 2026 that collectively reduce barriers across the<br>critical pipeline stages: person segmentation (MatAnyone2, 140MB model achieving state-of-the-art<br>video matting), inference acceleration (Diagonal Distillation, achieving 270x speedup over baseline<br>video generation), and mobile 3D rendering (MobileGS, achieving 120+ FPS Gaussian splatting on a<br>Snapdragon 8 Gen 3 phone at 4.8MB model size). We revise our feasibility classes accordingly, noting<br>that the "edge full video synthesis" class we originally projected at 12 months may now be achievable<br>within 6 to 9 months under moderate assumptions. We discuss implications for the policy<br>recommendations in our original analysis and argue that the structural mismatch between capability<br>maturation and regulatory response has widened.</p>
title Compressed Feasibility: A Technical Update to the Ambient Non-Consensual Synthesis Threat Model
topic Threat analysis
AI safety
deepfakes
perceptual consent
image-based sexual abuse
adversarial machine learning
augmented reality
AR wearables
Threat modeling
url https://doi.org/10.5281/zenodo.19242567