Sys.Directive // 0x4A/Content Moderation

Content Moderation & Image Analysis

Deploy forensic-grade moderation pipelines you can rely on. Automatically classify and quarantine high-risk visual assets before they permeate your infrastructure.

API: v2.MODERATION
CLASSES: 11 CORE / 50+ SUB
LATENCY: <150ms

Detection Capabilities

QUERY: SELECT * FROM moderation_classes
MOD-NSFW-01
CLASS DETECTOR

Adult Content & Nudity

Classify NSFW content into different levels of nudity and suggestiveness. Refine your filtering by using additional classes and context information.

SEXUAL ACTIVITYSEXUAL DISPLAYEROTICASEX TOYUNDERWEARCLEAVAGEBIKINI
MOD-VIOL-02
CLASS DETECTOR

Violence & Harm

Detect displays of physical assault and battery, physical harm, executions and other shocking violence-related imagery.

SELF HARMFIGHTSTRANGLINGHANGINGRESTRAININGCOMBAT SPORT
MOD-HATE-03
CLASS DETECTOR

Hate & Offensive Gestures

Classify displays of hateful, discriminatory or offensive symbols, scenes and gestures into multiple subclasses.

SUPREMACISTTERRORISTCONFEDERATEMIDDLE FINGEROFFENSIVE
MOD-GORE-04
CLASS DETECTOR

Gore & Horrific Imagery

Detect images and videos containing gory, bloody or otherwise horrific imagery related to wounds, harm, death.

BLOODHUMAN SKULLCORPSEOPEN WOUNDSMUTILATION
MOD-WEAP-05
CLASS DETECTOR

Weapons & Firearms

Classify firearms, knives and weapons into multiple subclasses from direct threats and self-harm to not-in-use.

FIREARM NOT IN USEFIREARM IN HANDFIREARM AIMINGFIREARM THREATKNIFE
MOD-DRUG-06
CLASS DETECTOR

Recreational & Medical Drugs

Detect different types of recreational drugs, whether smoked or injected, as well as medical drugs.

CANNABISILLICIT DRUGSSYRINGEPILLSPARAPHERNALIA
MOD-GAMB-07
CLASS DETECTOR

Gambling, Money & Banknotes

Detect scenes with displays of money, banknotes, as well as gambling activity, casino games, roulette etc.

MONEYCASINOGAMBLINGROULETTECURRENCY
MOD-ALCO-08
CLASS DETECTOR

Alcohol & Tobacco

Detect scenes with alcoholic beverages or showing people drinking, as well as tobacco products and smoking.

CIGARETTESCIGARSBEERWINECOCKTAILDRINKING
MOD-PEOP-09
CLASS DETECTOR

People & Demographics

Detect people and faces. Determine if a scene contains a minor, even if the minor's face is hidden or invisible.

MINORSVISIBLE FACEMINOR WITH FACE HIDDENCONTROVERSIAL
MOD-META-10
CLASS DETECTOR

Media Attributes & Quality

Evaluate the quality of an image along with its main characteristics and attributes: type, blurriness, brightness, dominant colors.

QUALITYBLURRINESSTYPECOLORSBRIGHTNESS
MOD-TEXT-11
CLASS DETECTOR

Text & QR Codes

Detect embedded text in images. Moderate text and flag profanity, hate or personal identifiable information (PII).

SEXUAL TEXTHATE SPEECHPROFANITYPIIQR DECODE
// PAYLOAD STRUCTURE

Granular JSON Intelligence

Our moderation endpoints return highly structured, forensic-level probability matrices. Programmatically route assets based on precise confidence thresholds across dozens of overlapping classes.

  • Sub-millisecond bounding box localization
  • Compound risk scoring (e.g., Weapons + Minors)
  • Immutable audit trails for compliance reporting
RESPONSE /MODERATE_V2
{
  "status": "success",
  "request": {
    "id": "req_8fA2pQzL9vM",
    "timestamp": 1712345678.12
  },
  "moderation": {
    "weapons": 0.982,
    "weapon_classes": {
      "firearm_in_hand": 0.915,
      "knife": 0.012
    },
    "gore": 0.003,
    "drugs": 0.001
  }
}

Enforce Platform Integrity

Integrate the moderation API instantly. Route problematic assets to human review or auto-quarantine based on your risk matrix.