SANDBOX
← All casesASI-TI-0003

[DEMO] TikTok: caption and comment risk signals

Evidence ReadymediumDemonstrationtiktokopened 29 Sep 2026 06:15 UTC

Synthetic demonstration case on TikTok. Research Tools exclude Pakistan, so live collection is blocked; the sandbox walk lists the exact dependency and the two lawful routes.

Demonstration case

Every item in this case is synthetic demonstration content. No live collection is connected in this environment, and no enforcement connection exists. Sandbox operation is not completed live platform coverage.
Source content

2 linked item(s)

  • Demonstrationtiktokvideolang: en

    [DEMO CONTENT] Fictional TikTok caption #demotrend #testonly — the video discusses a fictional dispute. Caption metadata only; spoken content is not searchable.

    https://www.tiktok.com/@aetnic.demo/video/7300000000000000001 · author @aetnic.demo

    Sandbox: TikTok Research Tools eligibility excludes Pakistan, so live collection is blocked pending a named collaboration with an eligible-region institution.

  • Demonstrationtiktokcommentlang: en

    [DEMO CONTENT] Fictional TikTok comment: "attack the convoy" appears as a synthetic test phrase. Illustrative string; no real event is referenced.

    https://www.tiktok.com/@aetnic.demo/video/7300000000000000001?comment=8880001 · author @demo.viewer.two

    Sandbox: comment text is reachable only via the Research API, which excludes Pakistan.

    • textSandboxconfidence 0.58

      Engaged 1 risk-term bucket: incitement language (1). Language: en. Review signal only.

      • 00:00–00:04attack the convoy (synthetic test phrase)

      Sandbox result: no live engine connected.

Each item shows its provenance badge and every analysis attached to it, including the engine that produced it.

Evidence

1 evidence item(s)

  • ev-tt-1Unsealedcontent hash verified against source
    Account ID
    tt-acct-0002
    URL
    https://www.tiktok.com/@aetnic.demo/video/7300000000000000001?comment=8880001
    TLD
    tiktok.com
    IP address(es)
    203.53.174.196
    Statutory provision
    PECA 2016 s.11 (hate speech) / s.37(1)
    Factual basis
    Synthetic demonstration basis: a comment contains an incitement test phrase. Written as fictional content for workflow demonstration.
    Captured at
    2026-09-29T06:15:00.000Z
    SHA-256
    73327c39d0bf733edac93d8fc38783f02030e0b82b7f3ad7b8ce6d202b464b9e
    Sealed at
    Not sealed
    Chain of custody
    • 29 Sep 2026 06:15 — M. F. Ashraf — Evidence item created from source content — Source item item-tt-2 on tiktok

Sealed evidence is immutable — the database rejects any later update or delete.

Agent assessment

1 agent run(s)

  • extractSandboxsandbox19 ms

    Extracted caption metadata and a comment test phrase. No media streams are available from TikTok.

    Sandbox — no Asterxa key configured

A failed run is reported as failed. The application never fabricates a result it did not get.

Decision trail

Approvals recorded

No approval decision recorded yet.

Identity or criticism of government alone can never justify enforcement — the server rejects it.

Outcomes

Recorded results

No outcome recorded yet.

Regulatory content is quoted for operational reference only and is not legal advice.

Rejected, failed and inconclusive are recorded distinctly and never collapsed.

Workflow

Case transitions

Only legal transitions are offered. Approval additionally requires a lawful ground recorded at the gate below.

Evidence

Assemble the PECA s.38 package

s.38 requires account ID, URL, TLD, IP address(es), the statutory provision and the factual basis. A capture timestamp, SHA-256 content hash and chain-of-custody log are added.

Agent operations

Run an agent on this case

Each run records the engine actually used. A sandbox run is a deterministic local analyser, never a live model.

Approval gate

Reviewer decision

Reviewer role required

You are acting as operator. Only a reviewer or admin may record an approval decision. The refusal is enforced on the server, not just hidden here.

Approval records a decision. It does not remove content, suspend an account, or guarantee that any platform acts.

Over-blocking risk: counter-speech, quotation, satire, journalism and newsworthy reporting are not violations. A lawful ground must be matched to a demonstrable content violation, not merely to the subject matter of the content.

Enforcement

Three separate tracks — never one action

Track A · Platform removal request

Asks the platform to remove the content. No platform grants a third party a programmatic removal API, so this resolves to a tracked manual escalation.

Track B · Account action

Would ask the platform to act against an account. This is never a power of this application and cannot be performed here.

Track C · Restriction within Pakistan

An exercise of Pakistani state power at the national network layer, via PTA. Available for all four platforms independently of platform cooperation.

Standing disclaimer

Restricts access INSIDE PAKISTAN. This is NOT removal: it does not delete content, does not touch the platform's servers, and does not affect anyone outside Pakistan. It is not hermetic — VPNs and encrypted DNS evade network-layer blocking. It is a power of PTA under PECA 2016 s.37(1) and the Removal and Blocking of Unlawful Online Content (Procedure, Oversight and Safeguards) Rules 2021 — NOT a power of this application. This application prepares and tracks a request; it never executes a block.

Each track carries its own honest capability flag. The application never submits a request it cannot lawfully complete.

Escalations

Tracked manual escalations

No escalations recorded for this case.

Where no verified integration exists, a named manual channel is opened and tracked rather than failing silently.

Outcome

Record the result — three distinct terminal states

Rejected, failed and inconclusive are different results. They are never collapsed into 'done'.