Preventing Unauthorized Adult and Age-Restricted Content Rendering to Children Through Hardware-Rooted Execution Finality
draft-das-child-safe-rendering-finality-01
This document is an Internet-Draft (I-D).
Anyone may submit an I-D to the IETF.
This I-D is not endorsed by the IETF and has no formal standing in the
IETF standards process.
The information below is for an old version of the document.
| Document | Type |
This is an older version of an Internet-Draft whose latest revision state is "Active".
|
|
|---|---|---|---|
| Author | Sangam Das | ||
| Last updated | 2026-08-27 | ||
| RFC stream | (None) | ||
| Formats | |||
| Stream | Stream state | (No stream defined) | |
| Consensus boilerplate | Unknown | ||
| RFC Editor Note | (None) | ||
| IESG | IESG state | I-D Exists | |
| Telechat date | (None) | ||
| Responsible AD | (None) | ||
| Send notices to | (None) |
draft-das-child-safe-rendering-finality-01
Network Working Group S. Das
Internet-Draft Independent Inventor
Intended status: Informational 27 August 2026
Expires: 28 February 2027
Preventing Unauthorized Adult and Age-Restricted Content Rendering to
Children Through Hardware-Rooted Execution Finality
draft-das-child-safe-rendering-finality-01
Abstract
Online child-safety controls commonly operate before the final
rendering boundary. Platforms may use account-age flags, parental
settings, content labels, recommender controls, server-side
classification, age-assurance systems, access policies, or
application filters to decide whether adult or age-restricted content
should be available to a user. Those controls are important, but an
upstream decision does not by itself guarantee that the content
cannot later be decrypted, decoded, composited, rendered, forwarded,
mirrored, or otherwise materialized through another software or
device path.
This problem is becoming more important as content delivery becomes
more distributed, encrypted, AI-mediated, personalized, and
dynamically generated. A modern device may receive content through
applications, browsers, content-delivery networks, embedded web
views, messaging clients, recommendation systems, generative-AI
services, caches, cloud gaming or streaming pipelines, local AI
models, or third-party SDKs. The security question is therefore no
longer only whether content was classified or whether an age check
occurred upstream. A later question must also be answered: is this
specific protected content authorized to become perceptible to this
recipient, on this device, under the current eligibility, policy, and
revocation state, at this moment?
This document defines a protected rendering execution-finality
architecture for adult, pornographic, sexually explicit, violent,
gambling-related, or otherwise age-restricted content. A proposed
rendering is represented as a Restricted Content Candidate Act and
remains in a Non-Renderable State until a Protected Enforcement
Domain validates the applicable recipient, content, device, policy,
age-or-eligibility, freshness, revocation, and sink predicates.
Protected validation evidence is committed before, or atomically
with, release of scoped non-bearer Rendering Finality Authority.
Das Expires 28 February 2027 [Page 1]
Internet-Draft Child-Safe Rendering Finality August 2026
A Protected Rendering Finality Sink independently verifies that
authority immediately before the content becomes perceptible.
Depending on the implementation, the sink may control content-key
release, decryption, media-decoder enablement, GPU or compositor
access, protected-surface creation, audio output, casting, screen
mirroring, display enablement, or an equivalent materialization
boundary. Content bytes may therefore be delivered to a device while
remaining technically non-renderable.
The architecture deliberately does not define a universal age-
estimation algorithm, identity system, or content-classification
scheme. Those mechanisms may supply inputs to the Protected
Enforcement Domain. This document defines the consequence-control
step that prevents an upstream policy result from becoming merely
advisory at the point of rendering.
UNICEF has warned that pornographic content can harm children and
that digital restrictions have not kept pace with technological
shifts. The ITU Child Online Protection programme provides global
guidance for safer digital environments, and the United Nations
Committee on the Rights of the Child has called for protection of
children from harmful content and online risks in the digital
environment. The European Commission has likewise adopted
protection-of-minors guidance and a privacy-preserving age-
verification approach for adult-restricted content. These materials
motivate the problem addressed here; they do not endorse this
particular technical architecture.
The central protocol principle is: permission to deliver content is
not permission to render it.
Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
This Internet-Draft will expire on 28 February 2027.
Das Expires 28 February 2027 [Page 2]
Internet-Draft Child-Safe Rendering Finality August 2026
Copyright Notice
Copyright (c) 2026 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents (https://trustee.ietf.org/
license-info) in effect on the date of publication of this document.
Please review these documents carefully, as they describe your rights
and restrictions with respect to this document. Code Components
extracted from this document must include Revised BSD License text as
described in Section 4.e of the Trust Legal Provisions and are
provided without warranty as described in the Revised BSD License.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 4
2. Why This Is Required Now . . . . . . . . . . . . . . . . . . 5
3. International Child-Safety Motivation . . . . . . . . . . . . 5
4. Core Security Distinction . . . . . . . . . . . . . . . . . . 6
5. Threat and Failure Model . . . . . . . . . . . . . . . . . . 6
6. Dual-Boundary Architecture . . . . . . . . . . . . . . . . . 7
7. Restricted Content Candidate Act . . . . . . . . . . . . . . 9
8. JSON Interoperability Profile . . . . . . . . . . . . . . . . 9
9. RestrictedContentCandidate JSON Schema . . . . . . . . . . . 9
10. RenderingPolicyDecision JSON Schema . . . . . . . . . . . . . 16
11. RenderingFinalityAuthority JSON Schema . . . . . . . . . . . 18
12. RenderSinkVerify Request JSON Schema . . . . . . . . . . . . 20
13. Protected Enforcement Domain Validation . . . . . . . . . . . 22
14. Protected Validation Evidence . . . . . . . . . . . . . . . . 22
15. Rendering Finality Authority . . . . . . . . . . . . . . . . 22
16. Where Enforcement Occurs . . . . . . . . . . . . . . . . . . 23
17. Content-Key and Decryption Enforcement . . . . . . . . . . . 23
18. Decoder, GPU and Compositor Enforcement . . . . . . . . . . . 23
19. Audio, Casting, Mirroring and Secondary Outputs . . . . . . . 24
20. Complete Allow Transaction Example . . . . . . . . . . . . . 24
21. Complete Denial Transaction Example . . . . . . . . . . . . . 26
22. Fail-Closed Requirements . . . . . . . . . . . . . . . . . . 27
23. Replay, Revocation and State Changes . . . . . . . . . . . . 28
24. Generated and AI-Transformed Content . . . . . . . . . . . . 28
25. Privacy Considerations . . . . . . . . . . . . . . . . . . . 28
26. Security Considerations . . . . . . . . . . . . . . . . . . . 29
27. Interoperability with Existing Age-Assurance and Platform
Systems . . . . . . . . . . . . . . . . . . . . . . . . . 29
28. Frequently Asked Questions . . . . . . . . . . . . . . . . . 29
28.1. FAQ 1 — How can Execution-Finality protect children
without creating continuous surveillance? . . . . . . . 29
Das Expires 28 February 2027 [Page 3]
Internet-Draft Child-Safe Rendering Finality August 2026
28.2. FAQ 2 — Does this require Apple, Google, Microsoft,
NVIDIA, AMD, Qualcomm, or other vendors to redesign their
hardware? . . . . . . . . . . . . . . . . . . . . . . . 31
28.3. FAQ 3 — Won’t cryptographic enforcement make video,
gaming, and AI rendering slow or battery-intensive? . . 32
28.4. FAQ 4 — Who issues and validates age credentials? . . . 33
28.5. FAQ 5 — Can VPNs, proxies, or alternative rendering paths
bypass it? . . . . . . . . . . . . . . . . . . . . . . 34
28.6. FAQ 6 — What happens when AI transforms restricted
content? . . . . . . . . . . . . . . . . . . . . . . . 35
28.7. FAQ 7 — How are Finality Leases protected against theft
and replay? . . . . . . . . . . . . . . . . . . . . . . 36
28.8. FAQ 8 — What happens if the operating system or
application is compromised? . . . . . . . . . . . . . . 37
28.9. FAQ 9 — Who controls the content policy? . . . . . . . . 38
28.10. FAQ 10 — Can governments or platforms abuse the mechanism
for censorship? . . . . . . . . . . . . . . . . . . . . 39
28.11. FAQ 11 — How does it work offline or during
authorization-service outages? . . . . . . . . . . . . 40
28.12. Central Architecture and Five Core Rules . . . . . . . . 40
29. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 42
30. Intellectual Property Considerations . . . . . . . . . . . . 42
31. Informative References . . . . . . . . . . . . . . . . . . . 42
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 43
1. Introduction
Child-safety systems frequently make policy decisions far from the
point at which content becomes visible or audible. A server may
classify content as adult-only. An account system may determine that
a user is under an applicable age threshold. A parental-control
service may deny a content category. A platform may enforce a
regional policy. Yet the media pipeline may contain many later
components: encrypted transport, local cache, decoder, GPU,
compositor, protected surface, display controller, speaker path,
casting interface, or screen-mirroring function.
The execution-finality model treats the final materialization of
restricted content as a separate consequence. The relevant security
property is not merely that a policy engine produced DENY. The
property is that the protected content cannot become perceptible
unless the final rendering boundary verifies current, act-specific
authority.
Das Expires 28 February 2027 [Page 4]
Internet-Draft Child-Safe Rendering Finality August 2026
The narrow scope of this document is adult and age-restricted content
rendering. It does not attempt to standardize child identity,
parental consent, age estimation, content moderation, recommender
algorithms, or the legal definition of harmful content across
jurisdictions.
2. Why This Is Required Now
The technical need has intensified because content is no longer
delivered through a single predictable application path. Streaming
platforms, browsers, encrypted media pipelines, social applications,
short-video feeds, generative AI, multimodal assistants, local
models, cloud rendering, game engines, mixed-reality systems, and
third-party SDKs can all participate in content selection and
materialization.
AI also changes the scale and dynamism of the problem. A system can
generate or transform sexually explicit or otherwise adult material
on demand, select content based on inferred interests, synthesize new
variants that were never pre-classified as static files, or route a
user through multiple content sources. An upstream catalogue label
alone is therefore less reliable as the sole technical enforcement
point.
This document does not assert that child-safety controls were
unnecessary in earlier Internet systems. The narrower claim is that
modern content stacks create more paths between upstream policy and
final perception, increasing the value of an independently verified
rendering boundary.
3. International Child-Safety Motivation
UNICEF states that pornographic content can harm children and has
expressed concern about the large quantity of pornography that is
easily accessible online. UNICEF has also stated that efforts to
restrict children's access in digital environments have not kept pace
with technological shifts.
The ITU Child Online Protection initiative provides policy-maker,
educator, industry, parent, and child-focused guidance intended to
support safer digital environments for children. Its current child
online protection work emphasizes child-centred and rights-based
protection mechanisms.
The United Nations Committee on the Rights of the Child, in General
Comment No. 25 (2021), treats the digital environment as encompassing
networks, content, services, applications, connected devices,
artificial intelligence, robotics, automated systems, algorithms, and
Das Expires 28 February 2027 [Page 5]
Internet-Draft Child-Safe Rendering Finality August 2026
data analytics. The Committee has repeatedly recommended protecting
children from harmful content and online risks while respecting their
rights and privacy.
The European Commission's protection-of-minors guidance under the
Digital Services Act addresses harmful content and recommends
effective age-assurance measures for adult content such as
pornography. The Commission's age-verification approach is designed
to let a user prove that an applicable age threshold is met without
unnecessarily revealing identity or exact age.
These sources provide policy and risk context only. They do not
define, require, or endorse the execution-finality mechanism
specified in this document.
4. Core Security Distinction
The central distinction is:
CONTENT DELIVERY
!=
CONTENT RENDERING AUTHORITY
AGE CHECK
!=
DISPLAY FINALITY
CONTENT CLASSIFICATION
!=
DECRYPTION AUTHORITY
SERVER-SIDE ALLOW
!=
DEVICE-SIDE MATERIALIZATION AUTHORITY
Figure 1
5. Threat and Failure Model
A restricted-content system may fail even when an upstream
classification or eligibility decision was correct. Examples include
stale age state, revoked parental policy, cache replay, alternate
application paths, direct decoder access, a compromised application,
a malicious embedded SDK, casting or mirroring, a stale decryption
key, sink substitution, modified content metadata, or a generated-
content path that bypasses the original classifier.
Das Expires 28 February 2027 [Page 6]
Internet-Draft Child-Safe Rendering Finality August 2026
A conforming deployment therefore assumes that ordinary application-
layer software may be buggy, compromised, overly permissive, or
unable to enforce the last consequence boundary by itself.
6. Dual-Boundary Architecture
A deployment MAY use two execution-finality boundaries.
The first boundary governs DELIVERY: whether protected content may be
transported to or retained within a device or protected environment.
The second boundary governs RENDERING: whether that content may
become visible, audible, cast, mirrored, decrypted into a usable
representation, or otherwise perceptible.
The second boundary is the critical boundary defined by this draft.
Content received after successful delivery remains non-renderable
until a current Rendering Finality Authority is independently
verified.
Das Expires 28 February 2027 [Page 7]
Internet-Draft Child-Safe Rendering Finality August 2026
Restricted Content
|
v
Encrypted / Protected Candidate Payload
|
v
Delivery Policy / Delivery PED
|
v
Delivery Finality Authority
|
v
Device Receives Protected Payload
|
v
STILL NON-RENDERABLE
|
v
Device-Side Protected Enforcement Domain
|
+-- recipient / eligibility state
+-- content class
+-- parental / regulatory policy
+-- device and session state
+-- policy / revocation epoch
+-- nonce / freshness
+-- intended rendering sink
|
v
Protected Validation Evidence
|
v
Scoped Non-Bearer Rendering Finality Authority
|
v
PROTECTED RENDERING FINALITY SINK
|
+-- PASS -> permitted rendering
|
`-- FAIL -> remain non-renderable
Figure 2
Das Expires 28 February 2027 [Page 8]
Internet-Draft Child-Safe Rendering Finality August 2026
7. Restricted Content Candidate Act
A Restricted Content Candidate Act represents a proposed
materialization of age-restricted content. It may describe a video
frame sequence, image, audio stream, interactive scene, generated
media object, XR scene, game asset, webpage media object, or another
protected representation.
The Candidate Act MUST remain non-renderable while required
execution-finality validation is incomplete.
8. JSON Interoperability Profile
This revision defines a JSON-based interoperability profile so that
implementations can evaluate a concrete contract rather than only an
abstract policy model. JSON objects MUST be encoded as UTF-8.
Implementations MUST reject unknown load-bearing fields when the
active schema declares additionalProperties=false.
The JSON profile does not require HTTP as the transport. The objects
may be carried over a local protected IPC mechanism, an operating-
system service boundary, a trusted application-to-enclave channel, a
platform API, or another authenticated transport. A future revision
may define a media type or explicit transport binding.
9. RestrictedContentCandidate JSON Schema
{
"$schema": "https://json-schema.org/draft/2020-12/schema",
"$id": "urn:ietf:params:json-schema:child-rendering-candidate:00",
"title": "RestrictedContentCandidate",
"type": "object",
"additionalProperties": false,
"required": [
"version",
"candidate_act_id",
"act_type",
"content",
"recipient_context",
"policy_context",
"render_request",
"freshness",
"finality_sink"
],
"properties": {
"version": {
"type": "string",
"const": "1.0"
Das Expires 28 February 2027 [Page 9]
Internet-Draft Child-Safe Rendering Finality August 2026
},
"candidate_act_id": {
"type": "string",
"minLength": 16,
"maxLength": 128
},
"act_type": {
"type": "string",
"enum": [
"RENDER_VIDEO",
"RENDER_IMAGE",
"PLAY_AUDIO",
"RENDER_INTERACTIVE",
"RENDER_GENERATED_MEDIA",
"CAST_PROTECTED_CONTENT",
"MIRROR_PROTECTED_CONTENT"
]
},
"content": {
"type": "object",
"additionalProperties": false,
"required": [
"content_id",
"content_digest",
"content_class",
"classification_source",
"protection_state"
],
"properties": {
"content_id": {
"type": "string",
"maxLength": 256
},
"content_digest": {
"type": "string",
"pattern": "^sha256:[0-9a-fA-F]{64}$"
},
"content_class": {
"type": "string",
"enum": [
"ADULT_SEXUAL_CONTENT",
"SEXUALLY_EXPLICIT_CONTENT",
"AGE_RESTRICTED_VIDEO",
"AGE_RESTRICTED_AUDIO",
"AGE_RESTRICTED_INTERACTIVE",
"AGE_RESTRICTED_GENERATED_MEDIA",
"OTHER_RESTRICTED"
]
Das Expires 28 February 2027 [Page 10]
Internet-Draft Child-Safe Rendering Finality August 2026
},
"classification_source": {
"type": "string",
"maxLength": 256
},
"classification_confidence": {
"type": "number",
"minimum": 0,
"maximum": 1
},
"protection_state": {
"type": "string",
"enum": [
"ENCRYPTED",
"KEY_WITHHELD",
"PROTECTED_SURFACE_ONLY",
"NON_RENDERABLE_BUFFER",
"OTHER_PROTECTED"
]
},
"origin_service": {
"type": "string",
"maxLength": 256
},
"generation_model_id": {
"type": [
"string",
"null"
],
"maxLength": 256
}
}
},
"recipient_context": {
"type": "object",
"additionalProperties": false,
"required": [
"recipient_context_id",
"eligibility_state",
"eligibility_evidence_ref"
],
"properties": {
"recipient_context_id": {
"type": "string",
"maxLength": 256
},
"eligibility_state": {
"type": "string",
Das Expires 28 February 2027 [Page 11]
Internet-Draft Child-Safe Rendering Finality August 2026
"enum": [
"ELIGIBLE",
"NOT_ELIGIBLE",
"UNKNOWN",
"EXPIRED",
"REVOKED"
]
},
"eligibility_evidence_ref": {
"type": "string",
"maxLength": 512
},
"age_threshold_claim": {
"type": [
"string",
"null"
],
"enum": [
"OVER_18",
"OVER_LOCAL_THRESHOLD",
null
]
},
"parental_policy_ref": {
"type": [
"string",
"null"
],
"maxLength": 512
},
"privacy_preserving_proof": {
"type": "boolean"
}
}
},
"policy_context": {
"type": "object",
"additionalProperties": false,
"required": [
"jurisdiction",
"policy_id",
"policy_epoch",
"revocation_epoch"
],
"properties": {
"jurisdiction": {
"type": "string",
"maxLength": 64
Das Expires 28 February 2027 [Page 12]
Internet-Draft Child-Safe Rendering Finality August 2026
},
"policy_id": {
"type": "string",
"maxLength": 256
},
"policy_epoch": {
"type": "integer",
"minimum": 0
},
"revocation_epoch": {
"type": "integer",
"minimum": 0
},
"parental_control_state": {
"type": "string",
"enum": [
"ALLOW",
"DENY",
"NOT_APPLICABLE",
"UNKNOWN"
]
},
"platform_policy_state": {
"type": "string",
"enum": [
"ALLOW",
"DENY",
"REVIEW",
"UNKNOWN"
]
}
}
},
"render_request": {
"type": "object",
"additionalProperties": false,
"required": [
"device_id",
"application_id",
"render_mode",
"output_targets"
],
"properties": {
"device_id": {
"type": "string",
"maxLength": 256
},
"application_id": {
Das Expires 28 February 2027 [Page 13]
Internet-Draft Child-Safe Rendering Finality August 2026
"type": "string",
"maxLength": 256
},
"component_id": {
"type": [
"string",
"null"
],
"maxLength": 256
},
"render_mode": {
"type": "string",
"enum": [
"LOCAL_DISPLAY",
"LOCAL_AUDIO",
"XR_DISPLAY",
"CAST",
"MIRROR",
"REMOTE_DISPLAY"
]
},
"output_targets": {
"type": "array",
"minItems": 1,
"items": {
"type": "string",
"enum": [
"DISPLAY",
"SPEAKER",
"HEADSET",
"XR_COMPOSITOR",
"CAST_SINK",
"MIRROR_SINK"
]
}
},
"requested_resolution": {
"type": [
"string",
"null"
],
"maxLength": 64
},
"requested_audio": {
"type": "boolean"
},
"requested_decryption": {
"type": "boolean"
Das Expires 28 February 2027 [Page 14]
Internet-Draft Child-Safe Rendering Finality August 2026
}
}
},
"freshness": {
"type": "object",
"additionalProperties": false,
"required": [
"nonce",
"created_at",
"expires_at"
],
"properties": {
"nonce": {
"type": "string",
"minLength": 16,
"maxLength": 256
},
"created_at": {
"type": "string",
"format": "date-time"
},
"expires_at": {
"type": "string",
"format": "date-time"
},
"session_id": {
"type": [
"string",
"null"
],
"maxLength": 256
}
}
},
"finality_sink": {
"type": "object",
"additionalProperties": false,
"required": [
"sink_id",
"sink_type"
],
"properties": {
"sink_id": {
"type": "string",
"maxLength": 256
},
"sink_type": {
"type": "string",
Das Expires 28 February 2027 [Page 15]
Internet-Draft Child-Safe Rendering Finality August 2026
"enum": [
"CONTENT_KEY_RELEASE",
"DECRYPTION_GATE",
"MEDIA_DECODER_GATE",
"GPU_COMPOSITOR_GATE",
"PROTECTED_SURFACE_GATE",
"DISPLAY_ENABLE_GATE",
"AUDIO_OUTPUT_GATE",
"CAST_GATE",
"MIRROR_GATE"
]
},
"attestation_ref": {
"type": [
"string",
"null"
],
"maxLength": 512
}
}
}
}
}
Figure 3
10. RenderingPolicyDecision JSON Schema
{
"$schema": "https://json-schema.org/draft/2020-12/schema",
"title": "RenderingPolicyDecision",
"type": "object",
"additionalProperties": false,
"required": [
"candidate_act_id",
"decision",
"decision_id",
"validated_policy_epoch",
"validated_revocation_epoch",
"permitted_effects",
"denial_reasons"
],
"properties": {
"candidate_act_id": {
"type": "string"
},
"decision": {
"type": "string",
Das Expires 28 February 2027 [Page 16]
Internet-Draft Child-Safe Rendering Finality August 2026
"enum": [
"ALLOW_RENDER",
"DENY_RENDER",
"ALLOW_RESTRICTED_MODE",
"REQUIRE_REVALIDATION"
]
},
"decision_id": {
"type": "string"
},
"validated_policy_epoch": {
"type": "integer",
"minimum": 0
},
"validated_revocation_epoch": {
"type": "integer",
"minimum": 0
},
"permitted_effects": {
"type": "array",
"items": {
"type": "string",
"enum": [
"RELEASE_CONTENT_KEY",
"DECRYPT",
"DECODE",
"CREATE_PROTECTED_SURFACE",
"COMPOSITE",
"DISPLAY",
"PLAY_AUDIO",
"CAST",
"MIRROR"
]
}
},
"denial_reasons": {
"type": "array",
"items": {
"type": "string",
"enum": [
"RECIPIENT_NOT_ELIGIBLE",
"ELIGIBILITY_UNKNOWN",
"ELIGIBILITY_EXPIRED",
"PARENTAL_POLICY_DENY",
"PLATFORM_POLICY_DENY",
"CONTENT_CLASS_MISMATCH",
"POLICY_EPOCH_STALE",
"REVOCATION_EPOCH_STALE",
Das Expires 28 February 2027 [Page 17]
Internet-Draft Child-Safe Rendering Finality August 2026
"NONCE_REPLAY",
"DEVICE_MISMATCH",
"APPLICATION_MISMATCH",
"SINK_MISMATCH",
"OUTPUT_TARGET_NOT_PERMITTED",
"ATTESTATION_FAILURE",
"AUTHORITY_EXPIRED"
]
}
},
"validation_evidence_ref": {
"type": "string"
},
"decision_expires_at": {
"type": "string",
"format": "date-time"
}
}
}
Figure 4
11. RenderingFinalityAuthority JSON Schema
{
"$schema": "https://json-schema.org/draft/2020-12/schema",
"title": "RenderingFinalityAuthority",
"type": "object",
"additionalProperties": false,
"required": [
"authority_id",
"candidate_act_id",
"candidate_digest",
"validation_evidence_ref",
"permitted_effects",
"sink_id",
"policy_epoch",
"revocation_epoch",
"nonce",
"issued_at",
"expires_at",
"consumption"
],
"properties": {
"authority_id": {
"type": "string"
},
"candidate_act_id": {
Das Expires 28 February 2027 [Page 18]
Internet-Draft Child-Safe Rendering Finality August 2026
"type": "string"
},
"candidate_digest": {
"type": "string",
"pattern": "^sha256:[0-9a-fA-F]{64}$"
},
"validation_evidence_ref": {
"type": "string"
},
"permitted_effects": {
"type": "array",
"minItems": 1,
"items": {
"type": "string"
}
},
"content_digest": {
"type": "string",
"pattern": "^sha256:[0-9a-fA-F]{64}$"
},
"recipient_context_id": {
"type": "string"
},
"device_id": {
"type": "string"
},
"application_id": {
"type": "string"
},
"sink_id": {
"type": "string"
},
"sink_type": {
"type": "string"
},
"policy_epoch": {
"type": "integer"
},
"revocation_epoch": {
"type": "integer"
},
"nonce": {
"type": "string"
},
"issued_at": {
"type": "string",
"format": "date-time"
},
Das Expires 28 February 2027 [Page 19]
Internet-Draft Child-Safe Rendering Finality August 2026
"expires_at": {
"type": "string",
"format": "date-time"
},
"consumption": {
"type": "string",
"enum": [
"SINGLE_USE",
"FRAME_WINDOW",
"SESSION_BOUND"
]
},
"max_frame_count": {
"type": [
"integer",
"null"
],
"minimum": 1
},
"max_duration_ms": {
"type": [
"integer",
"null"
],
"minimum": 1
},
"signature_or_mac": {
"type": "string"
}
}
}
Figure 5
12. RenderSinkVerify Request JSON Schema
{
"$schema": "https://json-schema.org/draft/2020-12/schema",
"title": "RenderSinkVerifyRequest",
"type": "object",
"additionalProperties": false,
"required": [
"candidate",
"authority",
"sink_runtime_state"
],
"properties": {
"candidate": {
Das Expires 28 February 2027 [Page 20]
Internet-Draft Child-Safe Rendering Finality August 2026
"$ref": "urn:ietf:params:json-schema:child-rendering-candidate:00"
},
"authority": {
"type": "object"
},
"sink_runtime_state": {
"type": "object",
"additionalProperties": false,
"required": [
"sink_id",
"current_policy_epoch",
"current_revocation_epoch",
"authority_consumed",
"device_attestation_state",
"output_path"
],
"properties": {
"sink_id": {
"type": "string"
},
"current_policy_epoch": {
"type": "integer"
},
"current_revocation_epoch": {
"type": "integer"
},
"authority_consumed": {
"type": "boolean"
},
"device_attestation_state": {
"type": "string",
"enum": [
"VALID",
"INVALID",
"UNKNOWN"
]
},
"output_path": {
"type": "string",
"enum": [
"DISPLAY",
"AUDIO",
"XR",
"CAST",
"MIRROR"
]
},
"screen_capture_active": {
Das Expires 28 February 2027 [Page 21]
Internet-Draft Child-Safe Rendering Finality August 2026
"type": "boolean"
},
"external_display_attached": {
"type": "boolean"
}
}
}
}
}
Figure 6
13. Protected Enforcement Domain Validation
The Protected Enforcement Domain (PED) MUST evaluate all predicates
required by the applicable deployment before releasing Rendering
Finality Authority. Typical predicates include recipient
eligibility, age-threshold evidence, parental policy, content
classification, content digest, jurisdiction, platform policy, device
identity, application identity, requested output mode, output target,
current policy epoch, current revocation epoch, nonce freshness,
device attestation, and intended Finality Sink identity.
A successful upstream age check MUST NOT itself cause decryption or
display. PED validation success authorizes creation of scoped
finality authority; it does not complete the rendering consequence.
14. Protected Validation Evidence
Before, or atomically with, Rendering Finality Authority issuance,
the PED MUST commit protected validation evidence. The evidence MAY
be represented by a protected signed record, MAC, enclave or HSM
assertion, sealed state, monotonic-state commitment, append-only
protected record, or equivalent protected mechanism.
A Boolean ALLOW value by itself is not sufficient finality authority.
15. Rendering Finality Authority
Rendering Finality Authority MUST be act-bound, content-bound,
recipient-bound, device-bound, application-bound where applicable,
policy-epoch-bound, revocation-epoch-bound, nonce/freshness-bound,
effect-bound, and sink-bound.
Possession alone MUST NOT be sufficient. The authority is non-bearer
because the Finality Sink independently verifies the protected
bindings before enabling the permitted materialization effects.
Das Expires 28 February 2027 [Page 22]
Internet-Draft Child-Safe Rendering Finality August 2026
An authority for DISPLAY MUST NOT automatically authorize CAST or
MIRROR. An authority for one device MUST NOT automatically authorize
another device. An authority for one content digest MUST NOT
authorize modified or substituted content.
16. Where Enforcement Occurs
The most important deployment question is where the final technical
gate is placed. The gate SHOULD control the first boundary at which
the protected content becomes perceptible or can be trivially
converted into a perceptible representation.
Possible Protected Rendering Finality Sinks include content-key
release, decryption, media-decoder enablement, GPU command
submission, protected-surface creation, compositor admission,
display-plane enablement, audio-output enablement, XR compositor
access, secure-video path enablement, casting, or screen mirroring.
A server-side policy endpoint alone is not a Protected Rendering
Finality Sink unless bypassing it also makes the protected rendering
technically impossible.
17. Content-Key and Decryption Enforcement
Where encrypted or otherwise protected media is used, withholding the
usable content key is a strong enforcement point. The key-release
component MAY act as the Finality Sink or as one element of a chained
finality path.
If validation fails, the usable key MUST NOT be released into an
unprotected application context. If a key was previously released
under a bounded authority, its validity SHOULD be constrained by
session, content, policy epoch, time, frame window, or equivalent
protected state.
18. Decoder, GPU and Compositor Enforcement
For media that is already locally encrypted or decoded inside a
protected pipeline, the Finality Sink MAY be placed at a decoder,
GPU, protected-surface, or compositor boundary. The implementation
MUST prevent ordinary application code from converting a denied
Candidate Act into a visible frame through an alternate unprotected
surface.
Protected rendering MAY therefore require that decoded frames remain
in protected memory or protected surfaces until the compositor
verifies current finality authority.
Das Expires 28 February 2027 [Page 23]
Internet-Draft Child-Safe Rendering Finality August 2026
19. Audio, Casting, Mirroring and Secondary Outputs
Adult or age-restricted content can be materialized through secondary
paths even when a local display is controlled. Audio, HDMI or
external displays, wireless casting, screen mirroring, remote-display
APIs, accessibility capture paths, and similar outputs MAY therefore
require separate consequence classes or sink-bound authority.
A local-display authority MUST NOT be interpreted as universal
authority for every secondary output.
20. Complete Allow Transaction Example
The following example shows an adult-classified video that remains
key-withheld until the device-side protected sink verifies current
eligibility, policy, device, content, freshness, and output-path
state.
{
"candidate": {
"version": "1.0",
"candidate_act_id": "rca-8f3d0d0d-20260826-000001",
"act_type": "RENDER_VIDEO",
"content": {
"content_id": "asset-adult-99182",
"content_digest": "sha256:aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",
"content_class": "ADULT_SEXUAL_CONTENT",
"classification_source": "platform-classifier-v7",
"classification_confidence": 0.998,
"protection_state": "KEY_WITHHELD",
"origin_service": "media.example",
"generation_model_id": null
},
"recipient_context": {
"recipient_context_id": "recipient-session-771",
"eligibility_state": "ELIGIBLE",
"eligibility_evidence_ref": "ageproof:opaque:44e2...",
"age_threshold_claim": "OVER_18",
"parental_policy_ref": null,
"privacy_preserving_proof": true
},
"policy_context": {
"jurisdiction": "EU",
"policy_id": "adult-content-policy-eu-v4",
"policy_epoch": 412,
"revocation_epoch": 39,
"parental_control_state": "NOT_APPLICABLE",
"platform_policy_state": "ALLOW"
Das Expires 28 February 2027 [Page 24]
Internet-Draft Child-Safe Rendering Finality August 2026
},
"render_request": {
"device_id": "device-3dd18",
"application_id": "app.media.client",
"component_id": "protected-player",
"render_mode": "LOCAL_DISPLAY",
"output_targets": [
"DISPLAY",
"SPEAKER"
],
"requested_resolution": "1080p",
"requested_audio": true,
"requested_decryption": true
},
"freshness": {
"nonce": "f9014acbe771440ab5208751",
"created_at": "2026-08-26T17:35:00Z",
"expires_at": "2026-08-26T17:35:30Z",
"session_id": "sess-fb211"
},
"finality_sink": {
"sink_id": "render-sink-secure-compositor-01",
"sink_type": "GPU_COMPOSITOR_GATE",
"attestation_ref": "attest:device-3dd18:epoch-88"
}
},
"authority": {
"authority_id": "rfa-d48d7f",
"candidate_act_id": "rca-8f3d0d0d-20260826-000001",
"candidate_digest": "sha256:bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb",
"validation_evidence_ref": "pve:9f1172",
"permitted_effects": [
"RELEASE_CONTENT_KEY",
"DECRYPT",
"DECODE",
"CREATE_PROTECTED_SURFACE",
"COMPOSITE",
"DISPLAY",
"PLAY_AUDIO"
],
"content_digest": "sha256:aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",
"recipient_context_id": "recipient-session-771",
"device_id": "device-3dd18",
"application_id": "app.media.client",
"sink_id": "render-sink-secure-compositor-01",
"sink_type": "GPU_COMPOSITOR_GATE",
"policy_epoch": 412,
"revocation_epoch": 39,
Das Expires 28 February 2027 [Page 25]
Internet-Draft Child-Safe Rendering Finality August 2026
"nonce": "f9014acbe771440ab5208751",
"issued_at": "2026-08-26T17:35:01Z",
"expires_at": "2026-08-26T17:35:30Z",
"consumption": "FRAME_WINDOW",
"max_frame_count": 900,
"max_duration_ms": 30000,
"signature_or_mac": "base64url:protected-authenticator"
},
"sink_runtime_state": {
"sink_id": "render-sink-secure-compositor-01",
"current_policy_epoch": 412,
"current_revocation_epoch": 39,
"authority_consumed": false,
"device_attestation_state": "VALID",
"output_path": "DISPLAY",
"screen_capture_active": false,
"external_display_attached": false
}
}
Figure 7
21. Complete Denial Transaction Example
The following example shows a denial where the recipient is not
eligible and parental policy also denies rendering. The important
result is not merely an error response: all materialization controls
remain disabled.
Das Expires 28 February 2027 [Page 26]
Internet-Draft Child-Safe Rendering Finality August 2026
{
"candidate_act_id": "rca-minor-session-223",
"decision": "DENY_RENDER",
"decision_id": "decision-99831",
"validated_policy_epoch": 412,
"validated_revocation_epoch": 39,
"permitted_effects": [],
"denial_reasons": [
"RECIPIENT_NOT_ELIGIBLE",
"PARENTAL_POLICY_DENY"
],
"validation_evidence_ref": "pve-deny:3882",
"decision_expires_at": "2026-08-26T17:36:00Z",
"sink_action": {
"release_content_key": false,
"decrypt": false,
"decode": false,
"create_protected_surface": false,
"display": false,
"play_audio": false,
"cast": false,
"mirror": false,
"result": "CONTENT_REMAINS_NON_RENDERABLE"
}
}
Figure 8
22. Fail-Closed Requirements
If required authority is missing, malformed, expired, stale,
replayed, revoked, consumed, content-mismatched, recipient-
mismatched, device-mismatched, application-mismatched, policy-
mismatched, output-mismatched, or sink-mismatched, the content MUST
remain non-renderable.
Failure MUST NOT be converted into a warning followed by display.
Implementations MUST NOT use 'best effort' rendering for content
subject to this protected finality profile when current authority
cannot be established.
Das Expires 28 February 2027 [Page 27]
Internet-Draft Child-Safe Rendering Finality August 2026
23. Replay, Revocation and State Changes
Eligibility and parental state may change after content is delivered.
A device may also restore an old application snapshot or replay a
previously valid rendering authority. The Finality Sink SHOULD
therefore verify current policy and revocation epochs immediately
before materialization and MUST enforce nonce or equivalent anti-
replay state.
An old authorization MUST NOT automatically override a later
parental-control change, account-state change, legal-policy change,
device revocation, or recipient eligibility change.
24. Generated and AI-Transformed Content
Generative AI may create age-restricted media that has no stable
catalogue identifier. The Candidate Act therefore supports content
digests, generation-model identifiers, classification sources, and
protected output state. A generated object SHOULD be bound to the
classification and authority actually used for the attempted
rendering.
This draft does not define how AI-generated content is classified.
It defines how a classification or policy decision can be made
technically load-bearing at the rendering boundary.
25. Privacy Considerations
Child-safety enforcement can itself create privacy risk if systems
unnecessarily expose a child's identity, exact age, browsing history,
or content choices. Implementations SHOULD minimize disclosure and
MAY consume privacy-preserving eligibility proofs that state only
whether the required threshold or policy condition is satisfied.
The rendering sink generally needs the validated eligibility result
and bindings necessary for effectuation; it does not necessarily need
the user's civil identity or exact date of birth.
Logs and validation evidence SHOULD avoid recording unnecessary
content titles, URLs, identity attributes, or detailed viewing
history where a digest, opaque identifier, or protected commitment is
sufficient.
Das Expires 28 February 2027 [Page 28]
Internet-Draft Child-Safe Rendering Finality August 2026
26. Security Considerations
Threats include compromised applications, forged age state, stale
parental policy, replayed authorities, content substitution after
classification, sink substitution, screen-capture bypass, casting
bypass, unprotected decoder paths, key extraction, rollback of
revocation state, and malicious SDKs.
Implementations SHOULD place load-bearing checks within protected
execution, trusted operating-system, hardware-backed, or equivalently
isolated enforcement paths when the threat model includes compromise
of ordinary application software.
27. Interoperability with Existing Age-Assurance and Platform Systems
The execution-finality layer is designed to consume outcomes from
existing systems rather than replace them. Inputs MAY come from
privacy-preserving age verification, platform account state,
parental-control systems, content classifiers, regulatory policy,
trusted identity providers, device policy, or enterprise/family
safety services.
Those systems answer whether a condition is satisfied. The Finality
Sink answers whether the specific protected rendering consequence may
now occur.
28. Frequently Asked Questions
28.1. FAQ 1 — How can Execution-Finality protect children without
creating continuous surveillance?
The biggest challenge is determining whether the person receiving
content is a child or an adult. A naive hardware-rendering gate
could require continuous camera access, facial analysis, or age
estimation. That would create serious privacy, security, and
regulatory concerns.
Execution-Finality should avoid continuous biometric surveillance
altogether. Instead, the architecture separates age assurance from
execution enforcement.
A trusted age-assurance provider can establish only the minimum fact
required, for example that the user belongs to an Under-18 or 18+ age
class. A privacy-preserving credential, potentially using zero-
knowledge techniques, can prove that condition without exposing the
user’s name, date of birth, face, or other unnecessary personal
information.
Das Expires 28 February 2027 [Page 29]
Internet-Draft Child-Safe Rendering Finality August 2026
Age Assurance
|
v
Privacy-Preserving Credential
|
v
Policy Decision
|
v
Finality Lease
|
v
Protected Rendering Sink
Figure 9
The hardware does not need to know who the person is. It only needs
to determine whether the current rendering operation possesses valid
authority under the applicable policy.
A short-lived Finality Lease can be issued for a particular device,
session, policy, and rendering destination. The lease can be
cryptographically bound to:
* age category or eligibility class
* device public key
* content or policy classification
* authorized rendering sink
* policy version
* expiration time
* nonce
* revocation information
* the protected session
This avoids creating a permanent adult token. Authorization can
instead be temporary, scoped, revocable, and bound to an
authenticated device or session.
Das Expires 28 February 2027 [Page 30]
Internet-Draft Child-Safe Rendering Finality August 2026
For a family device, the system can use explicit authorization
transitions. An adult can authenticate and establish an authorized
session; when the device is locked, handed over, or the authorization
expires, the protected session terminates. A child subsequently
using the device receives the applicable child policy.
Most importantly, age determination occurs upstream while finality
enforcement occurs downstream. The Finality Sink should not ask
whether the person is 17 or 18. It should ask whether this rendering
operation is currently authorized under the applicable policy.
Design principle: age should be an authorization attribute, not a
continuously monitored biometric state.
28.2. FAQ 2 — Does this require Apple, Google, Microsoft, NVIDIA, AMD,
Qualcomm, or other vendors to redesign their hardware?
No. That would be an unnecessarily difficult deployment model.
Execution-Finality should not prescribe a particular GPU, operating
system, processor, secure element, display controller, age-assurance
provider, or content-classification implementation. The standards
layer should define the protocol and security semantics of the
finality boundary.
Different platforms can implement that boundary differently. Apple
could use secure hardware and a protected media pipeline; Android
manufacturers could use trusted execution and their display stack;
Windows could integrate with protected media and graphics
infrastructure; smart TVs could use secure media pipelines; cloud-
gaming platforms could enforce the boundary in protected streaming
and rendering components.
Content / Application
|
v
Policy Engine
|
v
Finality Authority
|
v
OS Enforcement Point
|
v
Protected Media / GPU Path
|
v
Finality Sink
Das Expires 28 February 2027 [Page 31]
Internet-Draft Child-Safe Rendering Finality August 2026
Figure 10
The standards layer can define what constitutes valid execution
authority and how that authority must be verified, including
authorization evidence, cryptographic binding, content identity,
policy identifiers, device or sink identity, freshness, expiry,
revocation, replay protection, scope, and failure behavior. The
physical implementation remains platform-specific.
The adoption proposition is therefore not “build a completely new
child-safety processor.” It is “provide a trustworthy enforcement
boundary capable of validating standardized finality authority.”
A staged deployment path can be:
* Stage 1 — Software enforcement: browsers, applications, and
operating systems implement the protocol.
* Stage 2 — Trusted execution: authorization mechanisms move into
secure OS or TEE components.
* Stage 3 — Protected media paths: high-risk content is prevented
from bypassing the trusted execution path.
* Stage 4 — Hardware finality: new devices can place the final
enforcement point inside protected media, GPU, compositor,
display, audio, or XR pipelines.
* Stage 5 — Ecosystem adoption: content providers, AI providers,
browsers, app stores, and platforms use the standardized
interface.
The architecture is therefore hardware-rooted where hardware support
exists, but protocol-defined and implementation-neutral at the
standards layer. The interface is standardized; the implementation
is not.
28.3. FAQ 3 — Won’t cryptographic enforcement make video, gaming, and
AI rendering slow or battery-intensive?
It would if every frame had to undergo a complete cryptographic
authorization check. That should not be the design.
At 120 Hz, a display can process 120 frames every second. Performing
an expensive authorization operation on every frame would be
inefficient. Execution-Finality should therefore use authorization
epochs or Finality Leases.
Das Expires 28 February 2027 [Page 32]
Internet-Draft Child-Safe Rendering Finality August 2026
Authenticate
|
v
Verify Policy
|
v
Issue Finality Lease
|
v
Protected Session
|
v
Render Thousands of Frames
Figure 11
For example, a 10-minute lease could authorize a particular protected
session. Individual frames would not each need a new authorization
transaction. The lease can be cryptographically bound to Device +
Session + Content/Policy + Rendering Sink + Expiry + Nonce.
Another finality check occurs when a meaningful security state
changes, such as authorization expiry, device change, rendering-sink
change, content-classification change, policy change, revocation,
protected-session transfer, or a new high-risk content category.
For a 30-minute cloud-gaming session at 120 Hz, the architecture does
not perform 120 cryptographic authorization checks every second. A
session can be authorized once for a bounded epoch and then render
more than 200,000 frames under that protected session, subject to the
lease and revalidation rules.
The same principle applies to AI-generated content. An AI model can
generate large numbers of candidate outputs without individually
invoking the finality protocol. The protected effect boundary is the
critical point.
Architectural distinction: generation can be high-volume and
computationally cheap; effectuation must be authoritative and policy-
compliant.
Performance principle: cryptographically protect authorization state
transitions, not individual frames.
28.4. FAQ 4 — Who issues and validates age credentials?
This should not be controlled by a single global authority. The
architecture should support a federated age-assurance model.
Das Expires 28 February 2027 [Page 33]
Internet-Draft Child-Safe Rendering Finality August 2026
Possible credential issuers can include regulated age-assurance
providers, identity providers, government-approved providers,
platform-operated age-assurance services, and other trusted
authorities accepted by the relying party.
The important separation is: credential issuer is not necessarily the
policy authority, and neither is necessarily the rendering authority.
Age Provider
|
| “User satisfies 18+ condition”
v
Age Credential
|
v
Policy Authority
|
| “18+ may access this content”
v
Finality Authority
|
| scoped authorization
v
Finality Sink
Figure 12
The Finality Sink validates only the authorization evidence and
bindings required for the protected effect. A credential or
signature by itself is not execution authority. The executor must
correlate the evidence with the exact action or effect and make the
applicable authorization decision.
A stronger privacy design uses attribute-based credentials rather
than identity-heavy credentials. Instead of disclosing a person’s
identity and exact date of birth, the credential can establish only
that the authenticated subject satisfies the applicable age
threshold.
28.5. FAQ 5 — Can VPNs, proxies, or alternative rendering paths bypass
it?
A network-level mechanism can be bypassed. A protected effect
boundary is designed to address that weakness. A VPN can hide a
network destination; a proxy can alter the request path; alternative
DNS can bypass DNS filtering; and an AI model can transform content.
None of those should automatically grant authority to cross the
protected effect boundary.
Das Expires 28 February 2027 [Page 34]
Internet-Draft Child-Safe Rendering Finality August 2026
Untrusted Internet
|
VPN / Proxy / AI / Browser
|
v
Candidate Content
|
v
Policy Evaluation
|
v
Finality Authority
|
v
Finality Sink
|
v
Pixels
Figure 13
The important requirement is alternate-path closure. If an
application can simply bypass the protected rendering path and
directly access an unprotected display or output path, the
architecture is incomplete.
Every consequential rendering path that the deployment claims to
protect should converge on an authenticated enforcement boundary.
Depending on platform design, this may include the display
compositor, protected media decoder, GPU pipeline, XR compositor,
cloud-streaming client, secure display controller, audio path, cast
path, or mirror path.
The architecture does not claim to prevent physical capture. A
person can photograph a screen with another camera. Execution-
Finality protects the digital effect boundary, not physical reality.
28.6. FAQ 6 — What happens when AI transforms restricted content?
This is a critical issue because a restriction tied only to an
original URL or file identifier can disappear when an AI system
creates a modified representation.
Das Expires 28 February 2027 [Page 35]
Internet-Draft Child-Safe Rendering Finality August 2026
Restricted Image
|
v
AI Transformation
|
v
New Representation
|
v
Policy Re-evaluation / Authority Check
|
v
Finality Sink
Figure 14
The authorization should be associated not merely with a URL, but
with the material content or effect being authorized. Conceptually,
the finality decision can bind the original content digest, relevant
transformation lineage, final content digest, policy, age or
eligibility class, and destination sink.
The system does not necessarily need to understand every
transformation semantically. Low-risk transformations such as
resizing may preserve an existing authorization where policy permits,
while higher-risk generative modification can require a new policy
evaluation and new authority.
Key rule: transformation does not automatically create new authority.
An AI model can generate something new, but generation itself does
not make that output authorized. Computation is not authority.
28.7. FAQ 7 — How are Finality Leases protected against theft and
replay?
A Finality Lease should not be a simple bearer token. It should be
cryptographically bound to the conditions under which it was issued.
A lease can include or bind:
* lease identifier
* device public key
* session identifier
* content or policy digest
Das Expires 28 February 2027 [Page 36]
Internet-Draft Child-Safe Rendering Finality August 2026
* authorized sink identifier
* policy epoch
* revocation epoch
* expiration
* nonce
* permitted effect scope
The sink verifies the required bindings. Stealing the lease
therefore does not automatically allow an attacker to use it on
another device, sink, session, content item, policy epoch, or output
path.
Replay resistance can use unique nonces, expiration, session binding,
device-key binding, sink binding, policy epochs, atomic consumption
where appropriate, and revocation or status checks. Particularly
sensitive operations can use single-use authority.
A crucial distinction is that freshness is not the same as
authorization. A fresh lease can still be revoked or outside the
permitted policy. A valid signature proves integrity or control of a
key; it does not by itself prove that the protected effect is
authorized.
28.8. FAQ 8 — What happens if the operating system or application is
compromised?
The answer depends on the assurance profile. If everything is
controlled only by ordinary software and the operating system is
malicious, an attacker may bypass a software-only enforcement
mechanism.
For higher-assurance child-safety enforcement, the protected path can
be structured as:
Das Expires 28 February 2027 [Page 37]
Internet-Draft Child-Safe Rendering Finality August 2026
Application
|
v
Operating System
|
v
Trusted Execution / Protected Service
|
v
Protected Rendering Path
|
v
Hardware-Rooted Finality Sink
Figure 15
The application can be compromised without automatically gaining
finality authority. Stronger protection can use secure boot,
hardware-backed keys, trusted execution environments, authenticated
OS components, protected media paths, authenticated display or
compositor paths, and hardware-enforced access controls.
No architecture should claim complete protection if an attacker has
unrestricted control of the hardware. The standard can therefore
define assurance profiles such as Software Finality, OS/TEE Finality,
and Hardware Rendering Finality, allowing progressive deployment with
explicit security expectations.
28.9. FAQ 9 — Who controls the content policy?
The Execution-Finality protocol should not decide what is child-safe.
That is a critical separation. The architecture separates policy
creation, policy evaluation, and finality enforcement.
Policy Authority
|
v
Policy Identifier / Version
|
v
Policy Decision
|
v
Finality Authority
|
v
Finality Sink
Das Expires 28 February 2027 [Page 38]
Internet-Draft Child-Safe Rendering Finality August 2026
Figure 16
Different jurisdictions and organizations may legitimately apply
different policies. A national regulator, a school, a parent, a
platform, or another competent authority may define or select policy
under its own legal and operational context.
The protocol therefore carries or binds a policy identifier and
policy version rather than embedding one universal definition of
harmful or child-safe content. The Finality Sink enforces whether
the specific operation was authorized under the applicable policy; it
does not decide what society should consider appropriate.
28.10. FAQ 10 — Can governments or platforms abuse the mechanism for
censorship?
Yes. This is a legitimate architectural risk and should be addressed
explicitly. A mechanism capable of preventing one category of
content from reaching children could theoretically be expanded to
restrict other content or adult users.
Anti-abuse safeguards should include:
* Scope limitation: authority explicitly binds purpose, content
class, age or eligibility class, jurisdiction, policy version,
expiration, and authorized sink.
* Policy transparency: policies have identifiable versions, issuers,
scopes, and auditable changes.
* Separation of powers: age-credential issuance, policy definition,
finality-authority issuance, and hardware control should not
automatically belong to one entity.
* Privacy minimization: the Finality Sink should learn only what is
necessary to decide authorization, rather than collecting detailed
viewing histories or civil identity.
* Adult-rights separation: a child-safety profile should not
automatically become a general adult content-control profile.
These safeguards are important for interoperability, legitimacy, and
standards acceptance because the enforcement mechanism should be
technically narrow even when policy sources differ.
Das Expires 28 February 2027 [Page 39]
Internet-Draft Child-Safe Rendering Finality August 2026
28.11. FAQ 11 — How does it work offline or during authorization-
service outages?
The system should not require an online cryptographic transaction for
every frame or every media object. Short-lived Finality Leases allow
bounded offline operation after an authorization decision has been
made.
Online Authorization
|
v
Policy Evaluation
|
v
Finality Lease
|
v
Bounded Offline Operation
|
v
Lease Expiry
|
v
Renewal / Revalidation
Figure 17
Failure behavior should be risk-based and policy-defined. High-risk
content can fail closed when current authorization cannot be
established. Already-authorized lower-risk sessions may continue
only until the bounded lease expires, depending on policy.
Lease duration can vary by risk. Lower-risk operations may receive
longer leases; higher-risk operations may receive short leases or
require online authorization. The exact durations should be policy-
defined rather than hard-coded into the protocol.
This prevents a cached authorization from becoming a permanent bypass
while avoiding an online transaction for every rendered frame.
28.12. Central Architecture and Five Core Rules
Taken together, the FAQ answers produce the following child-safety
architecture:
Das Expires 28 February 2027 [Page 40]
Internet-Draft Child-Safe Rendering Finality August 2026
AGE ASSURANCE
|
v
Privacy-Preserving Proof
|
v
POLICY AUTHORITY
|
v
Policy Decision
|
v
FINALITY AUTHORITY
|
v
Short-Lived, Scoped Lease
|
+-- Device Binding
+-- Session Binding
+-- Policy Binding
+-- Sink Binding
+-- Expiry / Nonce / Revocation
|
v
+---------------------------+
| Protected OS / TEE |
| Enforcement Boundary |
+-------------+-------------+
|
v
Protected Path
|
v
+---------------------------+
| FINALITY SINK |
| GPU / Decoder / |
| Compositor / Display / XR |
+-------------+-------------+
|
v
PIXELS
Figure 18
The five core rules are:
* No continuous biometric surveillance.
Das Expires 28 February 2027 [Page 41]
Internet-Draft Child-Safe Rendering Finality August 2026
* No age credential is itself execution authority.
* No valid finality authority means no protected effect.
* No alternate path may bypass the protected effect boundary.
* Cryptographic verification protects authorization transitions, not
every frame.
29. IANA Considerations
This version requests no IANA actions. If a future version
standardizes an Internet media type, URI scheme, error registry, or
globally interoperable consequence-class registry, the corresponding
IANA considerations will be added.
30. Intellectual Property Considerations
Certain concepts described in this document are related to pending
patent applications in the DAS Protocols family. Any disclosure
required by an applicable standards-development process should be
made separately in accordance with that process. This section is
informational and does not specify licensing terms.
31. Informative References
[UNICEF-PORN]
UNICEF, "Protection of children from the harmful impacts
of pornography",
<https://www.unicef.org/harmful-content-online>. UNICEF
states that pornographic content can harm children and
that digital restrictions have not kept pace with
technological shifts.
[ITU-COP] International Telecommunication Union, "Child Online
Protection", <https://www.itu.int/en/ITU-
D/Cybersecurity/Pages/COP/COP.aspx>. ITU Child Online
Protection programme and guidelines.
[UN-CRC-GC25]
United Nations Committee on the Rights of the Child,
"General Comment No. 25 (2021) on children's rights in
relation to the digital environment",
<https://docstore.ohchr.org/SelfServices/
FilesHandler.ashx?enc=3ik0d0T8SX%2FpP07DX0nLHD%2Baw%2BVH0KHJ3a1Msix3LkRIphZtOKg5ykXnYtA79MjKsj%2Fql4vDUtT77VibA3Lp4g%3D%3D>.
United Nations guidance on children's rights in the
digital environment.
Das Expires 28 February 2027 [Page 42]
Internet-Draft Child-Safe Rendering Finality August 2026
[EU-MINORS]
European Commission, "Guidelines on the protection of
minors under the Digital Services Act", <https://digital-
strategy.ec.europa.eu/en/library/commission-publishes-
guidelines-protection-minors>. European Commission
protection-of-minors guidance.
[EU-AGE] European Commission, "The EU approach to age
verification", <https://digital-
strategy.ec.europa.eu/en/policies/eu-age-verification>.
Privacy-preserving age-verification approach for adult-
restricted online content.
Author's Address
Sangam Das
Independent Inventor
Balasore 756001
Odisha
India
Email: info@sangamdas.com
Das Expires 28 February 2027 [Page 43]