As AI’s footprint grows, facial data and image consent come under pressure
Pew’s web measurement, an exposed face-search archive, and nonconsensual alterations spreading on X place provenance, data custody, and image consent alongside content scale as growing online problems.
Artificial Intelligence··Midday
Pew measured AI signals on 10 percent of English pages
Pew Research Center ran about 490,000 English-language pages from Common Crawl through the Open Pangram detector. In a sample of 10,000 pages from July 2026, 10 percent showed significant signs of AI authorship or heavy editing; among pages published after ChatGPT’s release, the share exceeded one-third. The distribution differed by domain: roughly 10 percent on .com pages, 4.6 percent on .org, and about 1 percent on .edu and .gov. Pew explicitly notes that the detector can mislabel human writing and miss AI text, so the result describes a pattern across a large collection rather than settling the origin of any single page.[1]
A face-search archive left about 9 million files exposed
Security researcher Jeremiah Fowler found roughly 450 gigabytes of images in an unsecured Amazon S3 bucket reachable through an address in ClarityCheck’s public website code. Ars Technica reports that folders named “faces” and “profiles” held about 9 million files, while a second misconfiguration exposed email addresses and phone numbers. The images included adults, teenagers, and children, and people whose faces had been uploaded would usually have no way to know. The company closed the bucket after being contacted in July but disputed describing it as exposed, arguing that access required a specific, unindexed address.[2]
Small alterations preserve recognition while removing consent
In examples that 404 Media calls “subtlefakes,” real photographs of public figures are altered to look more revealing while remaining recognizable; some images carry the watermark of X’s own model, Grok. Actor Xochitl Gomez posted her parking-lot and red-carpet photographs beside altered versions on Instagram. The report says some accounts monetize both these slight alterations and openly nonconsensual images, while many accounts draw millions of views. X did not respond to 404 Media’s request for comment.[3]
