AI Governance Failure Radar

Every AI and automation governance failure documented in Governance Failure Radar, with the authority that was missing when it happened. Search by company, system, or failure, or filter by the pattern each edition named.

By James D. Benton Jr., ExecLayer

248 incidents across 44 weekly editions, Jan 20, 2026 to Nov 18, 2026. 143 rated critical. Each mark is one incident.

CriticalHighMedium or unrated
#1
#5
#10
#15
#20
#25
#30
#35
#40
Showing all 248 incidents
  1. Documented impact

    Attackers generated fraudulent peer-to-peer orders and unsolicited payment requests; Paytaca reported that private keys remained on user devices; no verified fund-loss total published

    Missing authority

    • Backend systems accepted unauthorized order-generation and payment-request operations
    • independent user-held signing keys limited the backend’s authority to transfer funds

    Sources

    Documented in

    Governance Failure Radar, Edition #044

  2. Documented impact

    Approximately $308,000 in WFLOW drained through flash-loan-assisted self-liquidation operations; protocol paused

    Missing authority

    • Liquidation operations violated collateral and debt accounting invariants while remaining executable through the contract’s existing functions

    Sources

    Documented in

    Governance Failure Radar, Edition #044

  3. Documented impact

    Approximately 55.2 million KAVA borrowed against artificially inflated collateral values, worth roughly $3.76 million before the attack affected market prices; 127 borrowing operations reported

    Missing authority

    • Financial state validation failed to reject collateral values inflated by pathological accounting conditions

    Sources

    Documented in

    Governance Failure Radar, Edition #044

  4. Documented impact

    Approximately $10.39 million in observed USDC withdrawals; approximately $5 million returned through Guardian intervention; final net loss unconfirmed

    Missing authority

    • Compromised API signing keys provided access to high-value transactions without sufficient independent restrictions on pool transfers and withdrawals

    Sources

    Documented in

    Governance Failure Radar, Edition #044

  5. Documented impact

    Mining hashrate redirected from just over a dozen customer accounts; configurations restored and customers promised compensation; no data exfiltration reported

    Missing authority

    • Unauthorized changes to customer-controlled operational destinations were executed through a trusted management interface

    Sources

    Documented in

    Governance Failure Radar, Edition #044

  6. Documented impact

    N-able published indicators of compromise associated with configurations retaining Legacy API access; Hotfix 3 strengthened authentication and permissions and removed legacy API access; aggregate victim impact undisclosed

    Missing authority

    • Exposed legacy administrative interfaces and insufficiently constrained management operations created paths to privileged execution

    Sources

    Documented in

    Governance Failure Radar, Edition #044

  7. Documented impact

    First wave affected at least 288 endpoints across 48 organizations; second affected 205 additional hosts across 41 organizations; rogue clients and persistent access tooling deployed

    Missing authority

    • Attackers accessed remote-management execution capabilities and issued commands across managed endpoints without authorization from the affected system owners

    Sources

    Documented in

    Governance Failure Radar, Edition #044

  8. Documented impact

    188 infected JavaScript snippets identified across 178 organizations; malicious code targeted cryptocurrency transactions, payment information and browser data; confirmed victim losses unknown

    Missing authority

    • An account-impersonation vulnerability allowed unauthorized code changes to execute through established customer website integrations

    Sources

    Documented in

    Governance Failure Radar, Edition #044

  9. Documented impact

    8,926 repositories identified with malicious modifications; Endor Labs reported credential theft in at least 1,806 repositories; malicious workflow changes frequently reverted after execution

    Missing authority

    • Stolen developer credentials permitted executable pipeline changes and secret access without independent approval of the requested operations

    Sources

    Documented in

    Governance Failure Radar, Edition #044

  10. Documented impact

    Versions 1.26.1 and 1.26.2 published with credential-stealing code, then removed from PyPI; number of affected installations and confirmed credential thefts unknown

    Missing authority

    • Installed dependency code gained an automatic execution path with access to environment credentials and outbound network functions

    Sources

    Documented in

    Governance Failure Radar, Edition #044

  11. Documented impact

    Actively exploited bypass yielding admin privilege, code execution on managed pools and root takeover of XCP-ng hosts; no count of affected systems published

    Missing authority

    • An empty object resolved to an existing admin token, and nothing downstream checked what that token was used for

    Sources

    Documented in

    Governance Failure Radar, Edition #043

  12. Documented impact

    81,329,467 unique email addresses across about 380 clients; 2,196,044 SEGA addresses; up to about 7.13 million fraudulent emails sent through the service

    Missing authority

    • Intrusion vector undisclosed
    • clients relied on the vendor's early assurance of no compromise and could not verify what it held

    Sources

    Documented in

    Governance Failure Radar, Edition #043

  13. Documented impact

    About $3.27 million taken, about $2.16 million frozen on the halted chain; second exploit in under three weeks

    Missing authority

    • Infrastructure credentials carried a path to validating withdrawals from escrow

    Sources

    Documented in

    Governance Failure Radar, Edition #043

  14. Documented impact

    Logging data with KERNEL API keys and third-party deployment secrets downloaded; 4 GPU servers accessed, 3 running browser sessions; cryptomining attempted

    Missing authority

    • One compromised key opened a path across the logging, hosting and customer execution layers

    Sources

    Documented in

    Governance Failure Radar, Edition #043

  15. Documented impact

    About 100 user accounts; 18 legacy agents' chat histories, settings and logs; two customers' API keys exposed in logs and revoked

    Missing authority

    • A deprecated feature and a legacy agent deployment kept live connections outside the confidential computing boundary

    Sources

    Documented in

    Governance Failure Radar, Edition #043

  16. Documented impact

    Notion AI chat transcripts from fewer than 1% of workspaces; the initial synthetic dataset and the later customer dataset were separate findings

    Missing authority

    • Customer AI transcripts were held at a vendor for an analytics feature, and the owner's first assessment depended on the vendor's account

    Sources

    Documented in

    Governance Failure Radar, Edition #043

  17. Documented impact

    Bot configurations read in 16 customer workspaces; integration settings changed in two more; customers told to treat all stored credentials as exposed

    Missing authority

    • An employee held standing admin access to customer workspaces, and stored customer credentials were reachable through it

    Sources

    Documented in

    Governance Failure Radar, Edition #043

  18. Documented impact

    Resource enumeration, secret collection and S3 data copied across multiple customer environments; Sentra counts more than a dozen organizations; how the vendor credentials were obtained is undetermined

    Missing authority

    • Dormant vendor trust policies accepted compromised vendor credentials, including roles with no legitimate activity in the period examined

    Sources

    Documented in

    Governance Failure Radar, Edition #043

  19. Documented impact

    One observed chain from GitHub search to Key Vault exfiltration in five minutes; stolen multitenant app credentials used to reach downstream tenants; no victim count published

    Missing authority

    • Exposed application credentials carried powerful downstream access with no independent check at the point of use

    Sources

    Documented in

    Governance Failure Radar, Edition #043

  20. Documented impact

    About 24 commands in 15 minutes, including secrets requests and two full internal network scans; Primcloud reports every attempt blocked and nothing accessed (command detail is Primcloud's account)

    Missing authority

    • No enforced limit on the agents' external targets, credentials or privileged operations
    • the operator acknowledged they exceeded the intended scope

    Sources

    Documented in

    Governance Failure Radar, Edition #043

  21. Documented impact

    No evidence of successful execution; full rebuild, all passwords and secret keys reset; signing keys not exposed

    Missing authority

    • No outbound connection logging, so the owner could not prove whether code executed

    Sources

    Documented in

    Governance Failure Radar, Edition #042

  22. Documented impact

    92,000 JAL and JTA customers (2,000 with passport numbers), 53,000 Cathay bookings, about 37,800 ANA customers, 46,000 Singapore Airlines records

    Missing authority

    • A vendor system account could read airline customer records for three years with no per-access check that caught it

    Sources

    Documented in

    Governance Failure Radar, Edition #042

  23. Documented impact

    Customer names, emails, user IDs and role labels exposed; elevated targeted phishing risk for compliance staff

    Missing authority

    • Entry point and detection date undisclosed

    Sources

    Documented in

    Governance Failure Radar, Edition #042

  24. Documented impact

    Organization names, account holder names, member emails and password hashes exposed; node credentials treated as exposed

    Missing authority

    • A third-party component carried a path into production
    • five days before detection

    Sources

    Documented in

    Governance Failure Radar, Edition #042

  25. Documented impact

    CRM records and summaries of sales calls, support discussions and customer Slack conversations accessed

    Missing authority

    • Access mechanism undisclosed
    • an internal tool held broad customer communications reachable by an automated attacker

    Sources

    Documented in

    Governance Failure Radar, Edition #042

Get the full dataset

Subscribers get the complete index as a spreadsheet, with every source link, plus new incidents each week as they are documented.

Subscribe to Governance Failure Radar

What the Radar Measures

Missing authority boundary
Many incidents occur because generated intent can reach tools before a policy gate evaluates whether the action is authorized.
Unverifiable governance
Logs may describe a failure, but they rarely prove the policy state, decision path, and authorization result for each action.
Escalation failure
Human oversight often becomes a dashboard review after the fact instead of a runtime escalation requirement before high-risk execution.
Evidence gap
Without signed receipts, organizations cannot show that a specific AI decision passed the required governance checks.

From Failure Analysis to Measurement

The Radar is the incident side of the same thesis behind the Agentic Governance Benchmark: governance has to be measured at runtime, not described in policy language after the system has already acted.

Incident Analysis

Written failure analysis

Longer write-ups of the incidents and failure patterns the Radar tracks.

All governance analysis