Abstract

An AI assistant begins reading a confidential financial summary through earphones. The user removes them, and the operating system switches to a room speaker. The first phrases have already been heard, a few more sit in a driver buffer, and the rest are still queued. The user was authenticated and the response was approved, yet nothing decides whether the remaining phrases may now be spoken to the room. Permission to generate or start a response is not permission to release every later part of it.

First public disclosure of the Mothership: PCT/IB2026/055615, filed 4 June 2026 and publicly disclosed 16 July 2026. Full text at Technical Disclosure Commons: https://www.tdcommons.org/dpubs_series/12542/ European public disclosure (Zenodo, 16 November 2026): “EUROPE’S NEXT AI SECURITY CHALLENGE: WHO CONTROLS THE FINAL ACTION? — From EU AI Act and NIS2 to 6G, the Digital Euro, Industrial AI and Critical Infrastructure.” https://zenodo.org/records/24274259

Based on Indian Provisional Filing: 202631136555, filed 15 November 2026, and disclosed with the Mothership (PCT/IB2026/055615) on 16 July 2026.

This patent-pending technical concept proposes continuing-authorization control for streamed, in-flight AI output such as speech, live transcription and AI summaries. The principle is that the actual action to control is the next effective disclosure, not the generation of the response. Authority to generate and authority to release each next unit are checked separately.

How it works

  1. The AI or application generates bounded units of output, and sensitive segments stay non-effective (staged) until released.
  2. The actual presentation is reconstructed: content class, selected device, real output route, queued units and presentation mode.
  3. Current authority is checked. It combines an audience contract, a current audience-state epoch, a current route epoch and revocation status.
  4. Release progress is resolved into three outcomes: committed (already past the controlled output boundary), uncertain (accepted by a device but not confirmed), and still releasable (held under an enforceable queue fence).
  5. Before each next bounded segment is released, its scope, audience, route, modality, epochs and sink are validated.
  6. A scoped presentation capability binds the permitted segment range, the destination and the protected output boundary.
  7. A defender-side Finality Sink releases only permitted units. On a route or audience change it stops controlled unreleased units and records the completed and uncertain output.

Already emitted sound or light cannot be recalled, and alternate outputs that bypass the gate are outside the guarantee. The gate sits at the actual compositor, audio path or receiving endpoint, and uses bounded checks per segment, not per audio sample or frame. This embodiment adds a time-sensitive question to the route question in Part 11: which unreleased units of an already running stream may continue when conditions change.

Prospective industry relevance (AI-generated; see notice below). In each mapping the gate sits at the controlled output boundary.

  • Apple (iOS, macOS, AirPods, AVAudioSession route changes): a privileged OS output authority builds the current presentation description and fences unreleased sensitive speech or notifications on a route change. Developer callbacks alone cannot revoke audio the hardware has already accepted.
  • Samsung (Android devices, Knox, multi-display): Knox or device-admin policy supplies content classification and current context. A privileged device implementation tracks release progress, fences buffers and mediates display and speaker release. This assumes vendor or OS participation, not an ordinary app.
  • Google (Android FLAG_SECURE, AudioManager device callbacks, Media3, Cast): sensitive response release is controlled across screen, speaker and casting endpoints. Stale output after a route change is stopped using the current epoch and queue progress. For remote Cast, a receiving-end release gate is needed.
  • Microsoft (Windows Media Foundation Protected Media Path, Teams, Windows desktop and conferencing): a supported protected media path is extended to separately authorized AI-generated segments and audience-state changes. The renderer or output path that controls dequeue acts as the gate. Locally private presentation and currently shared outputs get separate decisions.
  • OpenAI (Realtime API) and Anthropic (Claude Messages streaming, tool execution): a developer-owned client or tool dispatcher treats streamed audio and text chunks as uncommitted. It submits only authorized bounded units to a protected output broker. If the route changes or the session locks, unpresented sensitive chunks are fenced. Model generation does not establish what the speaker has physically emitted, and no native device-level gate in either product is claimed.
  • W3C Encrypted Media Extensions and other content-protection systems: the closest established mechanisms, and likely the strongest prior art an examiner would cite.

These mappings identify possible complementary integration points, not missing capabilities or deficiencies in the named products. Platform privacy controls, DRM, Protected Media Path and EME output restrictions are established and remain important. Whether the combination of release-progress accounting, dual-epoch authority and fenced reauthorization is novel is a patent-examination question and is not presumed. Practical assurances depend on endpoint integration, enforceable path coverage and measured testing.

Where to read the full filing: search “Sangam Das” or PCT/IB2026/055615 (pctib202655615) on WIPO PATENTSCOPE (https://patentscope.wipo.int/), or read the PCT file at Technical Disclosure Commons: https://www.tdcommons.org/dpubs_series/12542/

Basis and notice: The underlying invention is human-originated and based on the Mothership application PCT/IB2026/055615, filed 4 June 2026 and publicly disclosed 16 July 2026. This abstract and its industry and product comparisons were prepared with AI-assisted tools. They may contain errors, omissions, outdated facts or speculation. Readers must cross-check the original public PCT disclosure and the companies’ own documentation before drawing any conclusion. Company and product references are explanatory only. They imply no adoption, endorsement, validation, collaboration, licensing or infringement, and no finding on patent validity or third-party rights.

Part 12 of 39: this publication is Part 12 of a 39-part technical disclosure series, comprising 38 numbered embodiments and one separately designated Embodiment 12A.

License: The attached PDF is submitted under CC BY-NC 4.0 International, irrespective of the license displayed by the submission site.

KEYWORDS

AI Safety; Streaming Output Security; Real-Time AI Voice Security; Speech Output Privacy; Audio Route Security; Information Release Control; Media Content Protection; Protected Media Path; Digital Rights Management; Encrypted Media Extensions; Screen Sharing Data Loss Prevention; Multimodal Output Security; Agentic AI Security; AI Tool-Call Security; Consequence-Bound Authorization; Execution Finality; Finality Sink; Revocation; Receiver-Side Enforcement; EU AI Act; NIS2; Critical Infrastructure Protection; Mobile Security; Conferencing Security; Zero Trust; Low-Latency Enforcement; Legacy System Compatibility

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License

Share

COinS