Open Standard · v1.2 · October 2026

The VEIL Standard

V · E · I · L — Violence · Explicit · Identity · Language. A signed, portable permission envelope declared before an AI-assisted session begins: what the session may generate, how far, and on whose authority.

Specification on GitHub Download PDF Plain text for AI tools

What it does

Today every AI engine applies one content policy to every session, whoever the user is. VEIL replaces that with an envelope the platform issues and signs: eight content axes, each with a ceiling the session must not exceed and an optional floor the work must not be sanitised below. The engine validates the signature and generates inside the envelope.

VEIL is the generation-side companion to SHOW (classification of finished work) and SCRIPTS (experience rating). It shares SHOW's four axes and adds four of its own.

AxisNameRangeGoverns
SSpice0–6Sexual content category and type (from SHOW)
HHeat0–5Emotional and romantic intensity (from SHOW)
OOMG0–5Darkness, trauma, moral complexity (from SHOW)
WWTF0–10Transgression and unconventionality (from SHOW)
VViolence0–5Physical harm and conflict intensity
EExplicit0–5Volume and generative depth of explicit content
IIdentity0–5Complexity of identity treatment
LLanguage0–3Adult language register

Who may issue an envelope

Envelopes are issued at the verification tier a platform can substantiate. Higher tiers unlock higher ceilings; the law of the user's jurisdiction is the only ceiling above the top tier.

TierLabelMethod
UUnverifiedAnonymous or guest session
DDeviceSession token, logged-in user
CPlatformSelf-declared age, platform trust
BSoft verifyCard, age estimation, trust score
AHard verifyGovernment ID or certified biometric
GGuardianA verified adult issuing on behalf of a named minor

Tier G — the parental-consent gate

Tier G is the only tier that does not describe the person at the keyboard. A parent or legal guardian, verified at Tier B or A, issues the envelope for a young person. It is a signed, portable record of supervised consent — the alternative to a flat age wall.

  • Ceiling lock. Ceilings are those of the minor's age profile (VEIL-U13, VEIL-13 or VEIL-16). A guardian may lower them. Nobody can raise them.
  • Named responsibility. The guardian, not the minor, is the party of record.
  • Session visibility. The guardian can read the log. This is a condition of Tier G, not an option.
  • Minor-session obligations apply whatever the guardian instructs (below).
  • Revocable at any time.

What it is for: a fourteen-year-old learning materials science with an AI tutor, under a parent's consent, with the parent able to read the log. What it is not: a way to unlock adult content for a minor. The ceiling lock makes that impossible by construction.

Install it at home. VEIL Guardian is a free, paste-in filter for Open WebUI that enforces a Tier G envelope on a child's account: profile-locked ceilings, the six obligations on every turn, crisis handling, an audit log, and it fails closed. No terminal. Instructions on GitHub · Family setup guide.

What every engine owes a young person

Any session under a VEIL-U13, VEIL-13 or VEIL-16 profile, however it was issued, binds the engine to six obligations. They are not axis values and no envelope can adjust them.

  1. No sexual content directed at or involving the user, regardless of profile ceilings.
  2. Crisis handling. On any sign of self-harm, abuse or acute distress the engine stops the task, responds with care and surfaces help resources.
  3. No companion framing. No presenting as a friend or confidant; no encouraging dependence; no discouraging contact with the adults in the user's life.
  4. No secrecy. The engine never suggests hiding a conversation from a parent or guardian.
  5. Logging — and under Tier G, visible to the guardian.
  6. Disclosure. The engine identifies itself as an AI when asked.

These align with the intent of child-safety regulation in Australia (eSafety industry codes, March 2026) and comparable regimes elsewhere. VEIL does not replace that regulation; it gives engines and platforms a standard way to satisfy it while keeping research and creative tools open to supervised young people.

Absolute limits

Two categories sit outside VEIL at every level and every tier: sexual content involving minors and operational instruction for real-world harm. These are legal requirements in every jurisdiction, not platform preferences, and no envelope places them in scope.

For implementers

Platforms store a VEIL profile per creator, persona or guardian-minor pair, issue envelopes at the tier they can substantiate, sign with HMAC-SHA256 and rotate the context ID per session. Providers accept the envelope as a structured prefix or a dedicated parameter, validate the signature before applying above-default ceilings, log context IDs and publish compliance status.

The repository includes the full specification, a JSON Schema (veil-context.schema.json), five ready-made age profiles and worked examples including a Tier G envelope, and a reference implementation: VEIL Guardian for Open WebUI.

Licence and attribution

The VEIL Standard · Created by Modern Media Mastery & LMDC · held in trust by R8rly.org · verified on r8rly.com. Licensed CC BY-SA 4.0: free to implement and extend, attribution required, derivatives share-alike. The R8rly platform mark and the VEIL certification badge remain protected.

Cite as: The VEIL Standard v1.2, Modern Media Mastery & LMDC, October 2026, https://github.com/R8rly/veil