Abstract
Meta announced its Muse/Sentinel AI safety architecture on 8 September 2026, describing sandboxed agents that use surrogate credentials while real API keys and OAuth tokens remain protected outside the sandbox.
Interestingly, WIPO published international patent application WO 2026/150382 (PCT/IB2026/055615) on 16 July 2026, describing a closely related architecture, particularly in Embodiments 84 and 85.B2. Filed on 4 June 2026, the patent outlines host-side credential protection, opaque reference tokens, egress proxies and policy-based authorization.
This open industry validation exercise compares both architectures, examining technical similarities, differences and implementation details through patent disclosures, original drawings, terminology mapping and evidence-based analysis.
The objective is to encourage independent technical evaluation, transparency and industry discussion—not to allege copying or patent infringement. Architectural similarities do not establish derivation.
Full PCT filing and review guidance: https://zenodo.org/records/23728468
Primary Patent: WO 2026/150382 – WIPO
Earlier Related Publication: WO 2026/115520 – WIPO
Meta Comparison: How We Built Safety Into Muse, 8 September 2026.
Keywords: Meta Muse, Meta Sentinel, AI Agent Security, Credential Isolation, Agent Sandbox, Surrogate Credentials, Egress Proxy, API Security, Zero Trust, eBPF, Seccomp, Patent Prior Art, WIPO, WO 2026/150382.
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License
Recommended Citation
Das, Sangam Kumar, "An Open Industry Validation Exercise: AI Agent Credential Isolation in WO 2026/150382 and Meta Muse/Sentinel – A Comparative Analysis", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/12732