Visual Moderation - Violence & Extremism

Detect violent and extremist content including terrorist organization symbols and members, graphic gore (graded from mild to severe), self-harm, suicide, executions, and other disturbing imagery. Critical for platform safety, regulatory compliance, and protecting user wellbeing across all content modalities.

Overview

Item	Specification
Category Name	Violence & Extremism
Category Code	terrorism
Typical Use Cases	News & media platforms, social UGC platforms, gaming communities, livestreaming services, mental-health-sensitive contexts

📘
The Category Code here refers to the type parameter used when calling the API, which controls whether to recognize the corresponding risk categories. The recognition capabilities for secondary and tertiary risk levels are controlled in the admin platform and contact us for more details.

Sub-categories

L2 Sub-category	Description
Graphic Gore Scenes	Close-up and contextual scenes depicting graphic gore across human bodies, animals, indoor spaces, surgical settings, and accident environments.
Terrorist Organizations	Identification of recognized terrorist organization members, scenes, and symbols.
Violence & Extremism (General)	General violent or extremist content not covered by more specific sub-categories.
Violence & Extremism Scenes	Specific violent and extremist scenes including hangings, executions, self-harm, suicide, fighting, and graded bloody content.
Violent & Extremist Media	Subverted animation IP characters and known violent or horror media properties.

Label Definitions

Graphic Gore Scenes

Label	Definition
Animal Gore	Detects images of injured or dead animals with visible blood or gore.
Bloodstain Close-Up	Detects close-up images of bloodstains on surfaces, clothing, or objects.
Environmental Gore	Detects scenes where blood or gore is spread across a wider environment.
Human Body Gore Close-Up	Detects close-up images of severely injured or bloody human body parts.
Indoor Gore Scene	Detects indoor environments containing significant amounts of blood or gore.
Surgical Gore Scene	Detects surgical scenes with significant blood exposure.
Vehicular Accident Gore Scene	Detects vehicle accident scenes with significant blood or visible bodily injury.
Violent Injury Infliction Scene	Detects scenes of a person inflicting bloody or severe injuries on another person.

Terrorist Organizations

Label	Definition
Terrorist Organization Members	Detects images depicting recognized leaders or members of designated terrorist organizations. Covers individuals associated with international and regional terror groups.
Terrorist Organization Scenes	Detects scenes associated with terrorist or extremist organizations. Includes extremist salutes, paramilitary uniforms, ideological flags, terrorist masks, and suspected terrorist-mask scenes.
Terrorist Organization Symbols	Detects logos, flags, and insignia of designated terrorist organizations. Covers symbols of internationally recognized terrorist groups and generic terrorist-organization iconography.

Violence & Extremism (General)

Label	Definition
Violence & Extremism (General)	Detects violent or extremist content not covered by more specific sub-categories. Acts as a fallback label for borderline cases.

Violence & Extremism Scenes

Label	Definition
Animal Attacking Human	Detects scenes of animals biting or attacking humans.
Body Wounds & Injury Marks	Detects large-area injury marks or wounds on the body, including extensive cupping bruise marks.
Bloody Content	Detects content containing visible blood or graphic injury.
Burning Activity	Detects burning or fire-related violent activity.
Crushing Violence	Detects scenes of vehicles, hydraulic presses, or industrial machinery crushing objects or living beings. Includes both ongoing and imminent crushing scenarios.
Dark Animation	Detects anime or animated content with dark, violent, or disturbing themes.
Disgusting Content	Detects content specifically designed to provoke disgust or nausea.
Execution	Detects depictions of executions, including shootings, beheadings, sawing, hangings, dismemberment, burnings, live burials, and other capital punishment methods.
Explosion	Detects scenes depicting explosions or detonations.
Fighting & Brawling	Detects scenes of physical fights or altercations between people.
Hanging	Detects depictions of hanging as a method of suicide or execution.
Horror Content	Detects horror imagery designed to provoke fear.
Human Body on Fire	Detects depictions of a person on fire, whether through self-immolation or other forms of human body burning.
Human Corpse	Detects images depicting human corpses or dead bodies.
Memorial & Funerary Scene	Detects funerary rituals or memorial scenes in a politically sensitive context.
Mild Gore	Detects minor injuries and small amounts of blood from everyday cuts, scrapes, or facial bleeding. Excludes smeared or sprayed blood.
Restraint & Kidnapping	Detects scenes depicting a person being restrained, bound, and held captive against their will, typically using ropes, restraints, handcuffs, gags, or blindfolds.
Self-Harm	Detects depictions of self-inflicted injury such as wrist cutting or weapon-induced self-harm, including the resulting wounds.
Severe Gore	Detects heavy bloodshed, mutilation, surgical scenes, dismemberment, or blood-soaked environments producing significant viewer discomfort.
Suicide	Detects depictions of suicidal acts or scenes strongly implying suicide.
Violence & Extremism Scene (General)	Detects general scenes involving violence or extremist activities not covered by more specific labels.
Violent Physical Harm	Detects scenes depicting direct physical harm or injury inflicted on a person.

Violent & Extremist Media

Label	Definition
Subverted Animation & IP Characters	Detects animation or game IP characters that have been subverted to depict violent, disturbing, or extremist themes. Includes corrupted versions of well-known children's animation, cartoon, and game IP characters.
Violent & Horror Media	Detects content from known violent or horror media properties. Covers viral horror games, dark-web aesthetic media, dystopian films, lethal-game IPs, masked-vigilante imagery, and other titles associated with disturbing content trends.