Intelligence
Join the waitlist

Moona Intelligence

Live market intelligence for Agent Authority.

Here is what I am seeing.
No period over period comparison yet

I am not showing a period over period comparison yet, and I want to be direct about why. Moona Intelligence opened on 12 August 2026. A fair comparison needs two full windows collected the same way, and anything in the ledger dated before the desk opened is only there because a recent record cited it. That makes it a biased sample of its own period, not a baseline. The last 3 months comparison becomes available from 7 February 2027.

Knowledge checked 2026-10-27 13:19 UTC · Latest evidence 2026-10-01 10:13 UTC

Here is what I am seeing

Agent identity is where everyone is building. It is also where things keep breaking.

The market is treating agent identity as a control point. But 10 observed failures still map to this surface, so I am not convinced yet about where the winning implementation lands.

107 gap177 response10 failures64 organizations

How convinced I amHigh that it mattersModerate on where it lands

Why I read it that way

107 of the window's artifacts show that gap open in practice and 177 show a deliberate move against it, across 90 distinct developments. 64 separate organizations published on the response side: AWS, Algorand Foundation, AlpacaX, Ant Group, Anthropic and 59 others. High conviction that it matters, moderate conviction about where it lands.

Last 3 months, 4 July 2026 to 1 October 2026. 1282 independent artifacts, 163 showing a gap open and 194 showing a move against one.

Audit and Evidence carries more independent evidence than anything else I track this period, and it reads established. 163 of the artifacts in this window show a control gap open in practice, and 194 show somebody deliberately moving against one.

15 separate real world failures sit inside this window, every one of them a system doing what it had been permitted to do.

What I would do

2 things deserve attention right now
  • Now

    Go harder on approval controls.

    • 141 response artifacts against 59 showing the gap open.
    • 54 independent organizations have published on it in this window.
    • Three observed failures map to this surface.
    • High conviction on direction, accelerating.

    This is one of the cleanest directional signals on the board. 54 independent organizations behind it, and almost nothing failing under it.

    Why this matters for Moona

    What put it here

    • High conviction on direction, which is the level where I would build.
    • Response outweighs the gap, 141 against 59, so the market is answering this rather than only reporting it.
    • 54 independent organizations published on it, and no single publisher holds more than 9 percent of the response.

    What held it back

    • Three observed failures still sit on it.

    three failures sit against 141 response artifacts trying to close them, closer than it looks but still short of caught up.

    See the evidence
  • Watch

    Do not call audit and evidence yet.

    • 177 response artifacts against 149 showing the gap open.
    • 69 independent organizations have published on it in this window.
    • 12 observed failures map to this surface.
    • Low conviction on direction, accelerating.

    This is the surface I am least willing to act on, and not because it is quiet. It carries the largest evidence base on the board with the least resolved direction.

    What would resolve it

    What put it here

    • 69 independent organizations published on it, and no single publisher holds more than 10 percent of the response.
    • Accelerating inside this window.

    What held it back

    • 12 observed failures against 177 response artifacts, which is a heavier failure burden than the rest of the board carries.
    • Low conviction on direction. It clearly matters. I do not yet know where it lands.

    Do not buy a settled answer for audit and evidence yet: the response side edges it, 177 to 149.

    See the evidence
Why nothing else made the list
  • Execution Authority. Five observed failures against 151 response artifacts, which is a heavier failure burden than the rest of the board carries.
  • Delegated Authority. Eight observed failures against 156 response artifacts, which is a heavier failure burden than the rest of the board carries.
  • Agent Identity. 10 observed failures against 177 response artifacts, which is a heavier failure burden than the rest of the board carries.
  • Environment Boundaries. Eight observed failures against 156 response artifacts, which is a heavier failure burden than the rest of the board carries.
  • Human Oversight. Eight observed failures against 144 response artifacts, which is a heavier failure burden than the rest of the board carries.

What could change the call

  • Audit and Evidence keeps breaking while the market builds around it.

    12 separate real world failures in this window map to audit and evidence, against 177 artifacts showing a deliberate move against it. The response is running ahead of the failures on this surface, which is the shape of a problem being absorbed rather than solved. I would keep watching it either way. 15 failures sit in the window in total, so this is concentration rather than the whole picture.

    For now the response is growing faster than the failures. If that reverses, I change my read.

    See what sits behind it

  • Controls are arriving on agent identity and it still fails

    64 organizations published controls for agent identity inside this window, and 10 real world failures still landed on the same surface. Either the controls are not deployed where the failures happen, or they do not cover what breaks.

    Another failure on that surface with no matching control published would tell me the response is aimed at the wrong thing.

    See what sits behind it

This section exists to find the evidence that proves the call wrong.

Where I stand right now

Open a row for the evidence, the conviction and the records behind it

What would make me change my mind

  • Already in play

    Failures keep landing on agent identity while the controls keep shipping. Two more independent failures on that surface with no matching control published would tell me the response is not touching what actually breaks.

  • Not seen yet

    The response stops widening. If the 64 organizations currently publishing on agent identity stay the only ones, this is a vendor push rather than a market converging, and I would drop the conviction on direction.

  • Already in play

    Another surface takes the lead. Audit and Evidence is zero artifacts behind agent identity in this window. If that closes and holds, the call moves with it.

What I am watching next

Each trigger above is the same threshold the states themselves are computed from, so a reader can check the call rather than take it.

I could be wrong.

Signals I’m watching

  1. The response side on approval controls is already well ahead, 141 artifacts against 59, from 54 organizations that did not coordinate.Direction is resolved here, which is rare on this board.54 organizations · 838 artifacts · Accelerating
  2. 64 separate organizations published against agent identity in this window without coordinating.One of the largest evidence bases in this window, and the direction still has not resolved with it.64 organizations · 1073 artifacts · Accelerating
  3. Audit and evidence leans toward a response, 177 against 149, not by enough to call it settled.12 observed failures on one surface, more than any other carries in this window.69 organizations · 1136 artifacts · Accelerating

Three surfaces out of 8 on the board are carrying something worth elevating.

The evidence behind this

15Real world failuresRead for the control that was missing8Control surfaces characterisedOf 8 tracked. The rest hold too little evidence to call0Surfaces with counter evidenceThe market is building against the gap135Organizations moving controlsCounted on what they published46Research findingsPapers characterising the problem
Research in this window 46

Click any card or row to open the evidence behind it.

What arrived since the last update

Evidence dated 1 October 2026

One new record arrived, but none changed the thesis, directional conviction or what I would do.

See every record and what it did to my conviction

Where organizations are moving together

  • Agent IdentityConvergence forming64 organizations · 143 artifacts

    64 independent organizations published 143 artifacts against agent identity, with no single one behind more than 10 percent of it, and the response side now outweighs the gap 177 to 107. Read that as consensus forming rather than a launch cycle.

    AWS, Algorand Foundation, AlpacaX, Ant Group, Anthropic, Atlassian, AuditBadger, Australian Signals Directorate, Binance, Boomi, Britive, Broadcom, Cequence Security, ChainIT, Check Point Software Technologies, Cloudflare, Coder, CrowdStrike, Cursor, Cybersecurity and Infrastructure Security Agency, DataHub, Delinea, Docusign, Drata, European Commission, Fravity, GitHub, Google, HashiCorp, Hong Kong Monetary Authority, IBM, Improbable, JFrog, JetStream Security, LanternOps, Linux Foundation, Mambu, Mandatory Provident Fund Schemes Authority, Mate Security, Mezmo, Microsoft, MoonPay, NPCI, National Institute of Standards and Technology, Nuggets, Okta, OpenAI, Operant AI, Ping Identity, Rain, Reco, Rubrik, Rydoo, Salesforce, Seoul Labs, Socure, Solo.io, Somansa, SpaceXAI, University of Southern California, Visa, WRITER, Your Bourse, Zayo

  • Delegated AuthorityConvergence forming61 organizations · 127 artifacts

    61 independent organizations published 127 artifacts against delegated authority, with no single one behind more than 10 percent of it, and the response side now outweighs the gap 156 to 98. Read that as consensus forming rather than a launch cycle.

    AWS, Algorand Foundation, Ant Group, Anthropic, Atlassian, Australian Signals Directorate, Binance, Boomi, Britive, Broadcom, Cequence Security, ChainIT, Cheche Group Inc., Check Point Software Technologies, Cloudflare, Coder, CrowdStrike, Cursor, Cybersecurity and Infrastructure Security Agency, Delinea, Docker, Docusign, EMVCo, European Commission, GitHub, Google, HashiCorp, Hong Kong Monetary Authority, IBM, Improbable, JFrog, JetStream Security, LanternOps, Linux Foundation, Mambu, Mandatory Provident Fund Schemes Authority, Mate Security, Mezmo, Microsoft, MoonPay, NPCI, National Institute of Standards and Technology, Nuggets, Okta, OpenAI, Operant AI, Ping Identity, Rain, Reco, Rubrik, Rydoo, Salesforce, Seoul Labs, Solo.io, Somansa, SpaceXAI, University of Southern California, Visa, WRITER, Your Bourse, Zendesk

  • Execution AuthorityConvergence forming58 organizations · 125 artifacts

    58 independent organizations published 125 artifacts against execution authority, with no single one behind more than 10 percent of it, and the response side now outweighs the gap 151 to 78. Read that as consensus forming rather than a launch cycle.

    AWS, Algorand Foundation, Ant Group, Anthropic, Atlassian, AuditBadger, Australian Signals Directorate, Binance, Boomi, Britive, Broadcom, Cequence Security, ChainIT, Cheche Group Inc., Check Point Software Technologies, Cloudflare, Coder, CrowdStrike, Cursor, DataHub, Delinea, Docker, Docusign, Drata, Fortinet, Google, HashiCorp, Improbable, JFrog, JetStream Security, LanternOps, Linux Foundation, Mambu, Mezmo, Microsoft, MoonPay, NPCI, National Institute of Standards and Technology, Nuggets, Okta, Open Financial Technologies, OpenAI, Operant AI, Ping Identity, Rain, Reco, Rubrik, Rydoo, Salesforce, Seoul Labs, Solo.io, Somansa, SpaceXAI, TrueFoundry, University of Southern California, Virtue AI, WRITER, Your Bourse

  • Human OversightConvergence forming57 organizations · 121 artifacts

    57 independent organizations published 121 artifacts against human oversight, with no single one behind more than 11 percent of it, and the response side now outweighs the gap 144 to 95. Read that as consensus forming rather than a launch cycle.

    AWS, Ant Group, Anthropic, Atlassian, Australian Signals Directorate, Binance, Boomi, Britive, Broadcom, Cequence Security, Check Point Software Technologies, Cloudflare, Coder, CrowdStrike, Cursor, Delinea, Docusign, Drata, European Commission, Fortinet, Fravity, Google, HashiCorp, Hong Kong Monetary Authority, IBM, Improbable, JFrog, JetStream Security, LanternOps, Linux Foundation, Mambu, Mandatory Provident Fund Schemes Authority, Mate Security, Mezmo, Microsoft, MoonPay, NPCI, National Institute of Standards and Technology, Okta, OpenAI, Operant AI, Ping Identity, Rain, Reco, Rubrik, Rydoo, Salesforce, Seoul Labs, Socure, Solo.io, Somansa, SpaceXAI, University of Southern California, Virtue AI, Visa, WRITER, Zayo

  • Environment BoundariesConvergence forming56 organizations · 127 artifacts

    56 independent organizations published 127 artifacts against environment boundaries, with no single one behind more than 9 percent of it, and the response side now outweighs the gap 156 to 94. Read that as consensus forming rather than a launch cycle.

    AWS, Algorand Foundation, Ant Group, Anthropic, Atlassian, Australian Signals Directorate, Binance, Boomi, Britive, Broadcom, Cequence Security, Check Point Software Technologies, Cloudflare, Coder, CrowdStrike, Cursor, Cybersecurity and Infrastructure Security Agency, Delinea, Docker, Docusign, EMVCo, Google, HashiCorp, Hong Kong Monetary Authority, IBM, Improbable, JFrog, JetStream Security, LanternOps, Linux Foundation, Lovable, Mambu, Mandatory Provident Fund Schemes Authority, Mate Security, Mezmo, Microsoft, MoonPay, NPCI, National Institute of Standards and Technology, Okta, OpenAI, Operant AI, Ping Identity, Rain, Reco, Rydoo, Salesforce, Seoul Labs, Solo.io, Somansa, SpaceXAI, University of Southern California, Visa, WRITER, Your Bourse, Zayo

  • Approval ControlsConvergence forming54 organizations · 117 artifacts

    54 independent organizations published 117 artifacts against approval controls, with no single one behind more than 9 percent of it, and the response side now outweighs the gap 141 to 59. Read that as consensus forming rather than a launch cycle.

    AWS, Algorand Foundation, AlpacaX, Ant Group, Anthropic, Atlassian, AuditBadger, Australian Signals Directorate, Binance, Boomi, Britive, Broadcom, Cequence Security, ChainIT, Cloudflare, Coder, CrowdStrike, Cursor, DataHub, Delinea, Docker, Docusign, Fravity, GitHub, Google, HashiCorp, Improbable, JFrog, LanternOps, Linux Foundation, Mambu, Mezmo, Microsoft, MoonPay, NPCI, National Institute of Standards and Technology, Okta, Open Financial Technologies, OpenAI, Operant AI, Ping Identity, Rain, Reco, Rydoo, Salesforce, Seoul Labs, Socure, Solo.io, Somansa, SpaceXAI, University of Southern California, WRITER, Your Bourse, Zendesk

  • Sequence IntegrityConvergence forming44 organizations · 88 artifacts

    44 independent organizations published 88 artifacts against sequence integrity, with no single one behind more than 15 percent of it, and the response side now outweighs the gap 105 to 76. Read that as consensus forming rather than a launch cycle.

    AWS, Anthropic, Atlassian, Australian Signals Directorate, Boomi, Britive, Broadcom, Cequence Security, Check Point Software Technologies, Cloudflare, Coder, CrowdStrike, Cursor, Delinea, Docker, Docusign, European Commission, Google, HashiCorp, IBM, Improbable, JFrog, JetStream Security, LanternOps, Linux Foundation, Mambu, Mate Security, Mezmo, Microsoft, MoonPay, National Institute of Standards and Technology, OpenAI, Operant AI, Ping Identity, Rain, Reco, Rydoo, Seoul Labs, Solo.io, SpaceXAI, University of Southern California, Visa, WRITER, Zendesk

  • Audit and EvidenceBroad activity69 organizations · 145 artifacts

    69 independent organizations published against audit and evidence, which is broad activity. It is not yet convergence: the gap side still carries 149 artifacts against 177 on the response side, so the market is busy here rather than agreed.

    AWS, Algorand Foundation, AlpacaX, Ant Group, Anthropic, Atlassian, AuditBadger, Australian Signals Directorate, Binance, Boomi, Britive, Broadcom, Cequence Security, ChainIT, Check Point Software Technologies, Cloudflare, Coder, CrowdStrike, Cursor, Cybersecurity and Infrastructure Security Agency, DataHub, Delinea, Docker, Docusign, Drata, EMVCo, European Commission, Exiger, Fravity, Fullestop, Google, HashiCorp, Hong Kong Monetary Authority, IBM, Improbable, JFrog, JetStream Security, LanternOps, Linux Foundation, Mambu, Mandatory Provident Fund Schemes Authority, Mate Security, Mezmo, Microsoft, MoonPay, NPCI, National Institute of Standards and Technology, Nuggets, Okta, Open Financial Technologies, OpenAI, Operant AI, Penguin Ai, Ping Identity, Rain, Reco, Rubrik, Rydoo, Salesforce, Seoul Labs, Socure, Solo.io, Somansa, SpaceXAI, University of Southern California, Visa, WRITER, Your Bourse, Zendesk

Counted on who published, never on who was written about, and measured against how concentrated the publishing is.

What the evidence is about

Most of the evidence in this window sits in one cluster, Agent Authority, at 78 percent.

How the evidence is structured

Eight control surfaces carry enough independent evidence in this window to characterise.

Where companies are putting the control

  • GitHub210
  • AWS61
  • OpenAI61
  • IETF59
  • Anthropic48
  • Google40

121 organizations published separately against the same surface in this window, audit and evidence: AWS, Algorand Foundation, AlpacaX, Ant Group, Anthropic and 116 others.

What keeps breaking in the real world

Repeating: Audit and Evidence (12), Agent Identity (10), Delegated Authority (8), Environment Boundaries (8), Human Oversight (8), Execution Authority (5), Sequence Integrity (4), Approval Controls (3).

Explore every incident analysis

The thing I keep coming back to

Everyone wants autonomous agents right up until autonomy touches something that matters. Not one of the failures in this window needed a model to misbehave. The systems did what they had been permitted to do, which means the permission was the defect.

The evidence keeps landing in the same place. Audit and Evidence is where the decision gets made, and prompts are not policy at that boundary. The organizations that will come out of this well are the ones enforcing at that boundary and keeping evidence a reader could check afterwards.

1282 artifacts is what this rests on, with no previous window to measure it against. That is not a trend and I am not going to present it as one.

Explore by topic

Browse all topics

Latest Intelligence Records

See all Intelligence Records

The Moona Intelligence system

Four connected surfaces. Each one holds the evidence behind the one above it, and every link between them runs both ways.

  1. Intelligence

    What is changing in agent authority.

    You are here
  2. Records

    The verified evidence behind the analysis.

    228 published records
  3. Risks

    Evidence linked weaknesses and the vulnerabilities that instance them.

    34 weaknesses, 56 vulnerabilities
  4. Protocols

    How published architectures address those authority requirements, and where they do not.

    59 architectures assessed
Explore the evidence