Published October 6, 2026 | Version v1
Journal article Open

Ask HN: Technical Challenge — Can You Break This DAS vs Meta Muse/Sentinel Comparison? - Hacker News

Authors/Creators

Description

A Question for Readers

The relevant DAS disclosures were published from June 2026 by the World Intellectual Property Organization (WIPO), Geneva. Meta publicly disclosed Muse/Sentinel on 8 September 2026.

Readers are requested to independently identify which technical elements were already disclosed in the earlier DAS materials and which elements are genuinely Meta-specific engineering choices.

 

[ Recommended — Download the Complete Package ( ZIP FILE )  from the Download Section for Proper Evaluation ]

For Hacker News Readers

If you want to test the comparison yourself, download the complete source package:

00_COMPLETE_DAS_META_MUSE_SENTINEL_VALIDATION_PACKAGE.zip

For the fastest review:

Download ZIP → extract all files → place them in one AI Project/Workspace → read 00_READ_FIRST_AI_INSTRUCTIONS.md → reconstruct DAS first → reconstruct Meta Muse/Sentinel separately → only then compare them.

The supplied 01_AI_ANALYSIS_PROMPT.txt is optional. Use your own prompts too. Try to break the mapping rather than confirm it.

If you have only a few minutes, start with:

DAS I: Embodiment 84 + 85.B2
ALF-PCT: Embodiment 11

Then compare those passages directly with Meta’s source material.

The question for HN is:

Which correspondences survive a technically skeptical review, which disappear under scrutiny, and which Meta features are genuinely implementation-specific?

AI is only the research tool. The underlying primary sources are the evidence.

When Meta’s Muse/Sentinel architecture is compared with the earlier DAS disclosures, which technical elements were already disclosed before Meta’s 8 September 2026 announcement, and which are genuinely Meta-specific engineering choices?

Relevant DAS/ALF PCT materials were already publicly available through WIPO from June 2026, with relevant priority claims dating back to December 2025.

I am the author of DAS, so do not trust my conclusion. Download the sources, reconstruct DAS and Meta independently, then compare them yourself.

I am not alleging copying or infringement. Independent development is entirely possible. Chronology and technical correspondence are separate questions.

Focus on the actual control structure:

agent computation → authority decision → enforcement boundary → bounded permission → external effect

Then test the differences: sandboxing, egress control, credentials, human approval, provenance/taint, anti-bypass enforcement, and where the action becomes externally effective.

If you have only a few minutes, start with:

  • DAS I: Embodiment 84
  • DAS I: 85.B2 — Agent Sandbox Egress Finality Gate
  • ALF-PCT: Embodiment 11

Then compare them with Meta’s own Muse/Sentinel technical materials.

Use AI if useful, but AI is not the evidence. The evidence is the underlying primary-source record.

If the proposed mapping is wrong, identify exactly where the causal architectures diverge.

Do not accept my framing. Try to break it.

Underlying principle:

Computation is not authority to act.

#AI #AISecurity #AIAgents #Cybersecurity #MetaMuse #MetaSentinel #ExecutionFinality

Notes

After completing the source-by-source analysis, readers may wish to consider the following questions independently:

 Chronology and primary evidence:
What specific technical mechanisms were expressly disclosed in the DAS primary-source record before Meta publicly introduced Muse/Sentinel on 8 September 2026, and can each one be verified from dated patent/publication records?

Architecture, not terminology:
After ignoring names and terminology, do DAS and Muse/Sentinel implement the same or materially similar causal control structure—for example:
proposed action → independent authorization/validation → bounded permission → non-bypassable enforcement → external effect?

Component-by-component verification:
For every claimed correspondence, what exact DAS passage and what exact Meta source support the mapping, and is the correspondence expressly disclosed, necessarily implied, or merely inferred?

Differences and non-overlap:
What significant mechanisms exist only in DAS, what mechanisms exist only in Muse/Sentinel, and which differences are merely implementation choices—such as VM/container, proxy, eBPF/LSM, credential broker, worker isolation, TEE/HSM—or instead change the underlying authority model?

Alternative explanation and final conclusion:
Could the observed similarities reasonably result from independent engineering of the same technical problem or established prior art, and after considering both similarities and differences, what is the narrowest conclusion actually supported by the evidence?

 

 

Series information

Series information

Official WIPO PATENTSCOPE Records for Independent Cross-Verification

Readers are strongly encouraged to verify the underlying patent record directly through WIPO PATENTSCOPE rather than relying solely on this article, the mapping documents, or AI-generated analysis.

ALF PCT — PCT/IB2026/053385
International Publication: WO 2026/115520
Filed: 7 April 2026 · Published: 4 June 2026
Official WIPO PATENTSCOPE Record

THE DAS PROTOCOLS — PCT/IB2026/055615
International Publication: WO 2026/150382
Filed: 4 June 2026 · Published: 16 July 2026
Official WIPO PATENTSCOPE Record

THE DAS PROTOCOLS — Part II — PCT/IB2026/055760
International Publication: WO 2026/150383
Filed: 7 June 2026 · Published: 16 July 2026
Official WIPO PATENTSCOPE Record

THE DAS PROTOCOLS — Part III — PCT/IB2026/055870
International Publication: WO 2026/150384
Filed: 10 June 2026 · Published: 16 July 2026
Official WIPO PATENTSCOPE Record

DAS PROTOCOLS — Part IV — PCT/IB2026/056058
International Publication: WO 2026/154461
Filed: 13 June 2026 · Published: 23 July 2026
Official WIPO PATENTSCOPE Record

DAS PROTOCOLS — Part V — PCT/IB2026/056809
International Publication: WO 2026/172334
Filed: 1 July 2026 · Published: 20 August 2026
Official WIPO PATENTSCOPE Record

For any material technical conclusion, readers should cross-check the relevant passage against the corresponding WIPO publication and its dated international filing record.

For independent verification, readers should cross-check the claimed priority data against the corresponding published WIPO/PCT records and their listed priority documents.

 

 

Technical and Legal Scope Note

This article is limited to a technical discussion of publicly available materials, documented filing/publication chronology, and a reproducible primary-source validation methodology.

It does not allege or imply copying, misappropriation, access to unpublished material, infringement, inducement, ownership, derivation, lack of independent development, or any unlawful conduct by Meta or any other person or organisation.

References to filing dates, publications, Indian provisional applications, PCT applications, or claimed priority dates are provided solely to identify the documentary chronology. They should not be interpreted as a determination of patent priority, entitlement to priority, claim scope, validity, novelty, inventive step, enforceability, infringement, freedom to operate, or ownership. Those questions depend on the applicable law, the relevant claims, supporting disclosures and jurisdiction-specific legal analysis.

Any discussion of architectural correspondence is technical and functional only. Similar security objectives, control functions or architectural patterns do not by themselves establish copying, legal equivalence, patent coverage or infringement.

Readers are encouraged to review the cited primary sources independently, test both similarities and differences, and reach their own conclusions.

Files

1-PATENT-FILING-COPY-PCTIB2026-055615.pdf

Files (75.3 MB)