# Navigating the Identity Abyss in the AI-Native Era > [!question] Lingering Questions > - "ID jug" in the auto-captions is almost certainly ID-JAG — verify the actual standard's name and status before citing it anywhere. > - MCP EMA removes consent from the user entirely — what is the actual review and revocation surface? The talk doesn't say. > - The 500-consent-decisions figure is a cross-product worst case (10 MCP servers × 5 agents × 10 scopes), not a measurement of any real deployment. What's the real number? > [!tip] Article Ideas > - Read this alongside a joost.blog piece on the 2026-07-28 MCP spec release — this talk covers the identity side, joost.blog covers the server side of the same release. > [!link] Cross Reference > - [[3. Resources/Permanent Notes/Delegated Consent Reduces Burden by Moving Visibility Upstream]] — an existing Permanent note already captures this talk's consent-delegation tradeoff; this Literature note supplements it rather than duplicating it. > - [[3. Resources/MOCs/AI Agent Safety MOC]] > [!quote] Vocabulary > - **ID Numa** (Japanese) / **"identity abyss"** (the speaker's English rendering) — the experience of identity standards having proliferated past navigability; the talk's most portable contribution is naming a feeling practitioners already have. > - **MCP EMA (Enterprise Managed Authorization)** — an MCP extension that delegates agent consent decisions to the identity provider rather than surfacing each one to the end user. ## Bibliographic capsule A KubeCon Japan 2026 talk arguing that identity standards have proliferated past navigability, and that agentic AI makes this materially worse through combinatorial consent: the worked example is 10 MCP servers × 5 agents × 10 scopes = 500 consent decisions a user would otherwise have to make one by one. ## Author's thesis Identity standards were already hard before agents; agents and MCP servers being "dynamic and nested" makes complexity multiplicative rather than additive, and the practical response — MCP EMA delegating consent to the identity provider — trades user burden for user visibility without naming that it's doing so. ## Key arguments (≤3) 1. **Identity standards have become an abyss** — the talk's naming contribution, setting up a three-era progression from web app to API economy to agentic. 2. **Nested systems multiply authorization decisions** — agents calling servers that call further servers scale permission decisions as a product of the layers, not a sum. 3. **Removing a decision from the user relocates it rather than eliminating it** — delegated consent removes burden and visibility together; the question is who decides and whether the user can see it, which the talk doesn't fully address. ## Claim atoms ### Identity standards have become an abyss > [!quote] ID:rw_01kysr41xdxn700bt76cpb06s1 > In English, I want to call it as identity abyss. Mid-talk, introducing the Japanese term ID Numa and its English rendering. The naming is the talk's most portable contribution — a shared label for an experience practitioners already have. ## Permanent candidates - **Nested systems multiply authorization decisions** — when agents call servers that call further servers, permission decisions scale as a product of the layers rather than a sum. → promoted. - **Removing a decision from the user relocates it rather than eliminating it** — already exists as [[3. Resources/Permanent Notes/Delegated Consent Reduces Burden by Moving Visibility Upstream]]. Not recreated. ## Processing checklist - [ ] Verify the "ID jug" / ID-JAG standard name and status before citing - [ ] Check whether Keycloak's MCP EMA support has a documented consent review surface - [x] Connect into the relevant MOC ## Links - [[3. Resources/MOCs/AI Agent Safety MOC]] — agent authorization, consent delegation, and the safety-versus-visibility tradeoff are core to this MOC - [[3. Resources/Readwise/Full Document Contents/Articles/Navigating the Identity Abyss in the AI‑Native Era - Yuichi Nakamura, Hitachi]]