Your fact-checks will appear here
X's content moderation policies and enforcement actions, including revised Terms of Service and expanded content definitions
Grok’s Spicy Mode is not a simple “off switch” for but a distinct generation pathway that relaxes default filters, inserts additional creative processing steps, and layers supplemental safety checks—r...
Snapchat’s “My Eyes Only” is presented as an encrypted, passcode‑protected vault for Snaps moved from Memories and claims that content there is protected even if a device is stolen . However, Snapchat...
A video prompt can initially show “” and then appear after repeated attempts because ’s is conservative, context-sensitive, and applied at multiple stages—sometimes flagged late in generation—while en...
Platforms’ responses to have ranged from ad hoc to hurried creation of notice-and-removal systems under legal pressure, but enforcement remains uneven and contested; new U.S. law — the — now forces co...
It is reasonably likely that at least one xAI employee—or a content-moderation system tied to xAI—reviewed and flagged AI-generated images that were borderline (partial nudity, not overtly sexual), be...
PhotoDNA is a perceptual “fuzzy” hashing system developed by Microsoft and Dartmouth to create robust digital fingerprints of known images so platforms and law enforcement can find copies even after c...
frames as both an events platform and a site tightening legal and AI-related controls: it is actively promoting a 2026 marketing calendar and live-event features while simultaneously rolling out revis...
in 2025 converge on broad prohibitions against illegal, non-consensual, and high-risk and , but diverge sharply in allowable nuance: some (notably big-platform models) enforce tighter, conservative li...
look for technical markers ranging from low‑level pixel and spectral artifacts to high‑level biological and behavioral signals, and they often combine multimodal clues like metadata and liveness signa...
’s reporting portrayed ’s internal moderation apparatus as understaffed, ad hoc, and tolerant of “gray area” content choices, while employees described traumatic working conditions and privacy fears a...
AI companies generally follow a common playbook when they encounter child sexual abuse material (CSAM): detect with automated tools, remove and preserve evidence, and submit reports to the National Ce...
When unverified survivor testimony surfaces that names public figures, professional newsrooms balance legal constraints, ethical duties to survivors, and verification standards before publishing, whil...