Are our AI agents under control — at a glance, for the board.

One number, five categories, the open risks that matter. No operator tables — this is what leadership and the board need to read in two minutes.

Governance Index

72 / 100

+4 pts WoW

Needs attention

A blended score across egress control, data protection, identity, spend, and coverage. Up this week, but still below the 80 leadership target.

72 / 100

0target 80100

Posture By Category

Where we are strong, where we are exposed.

Egress Control

40default-allow

Data Protection

65needs work

Identity & Access

78on track

Spend Discipline

70watch

Coverage

82healthy

Egress control is the limiting factor — the org is default-allow, so any agent can reach any endpoint until policy is inverted.

Open High-Risk Findings

The three issues leadership should be tracking.

Customer PII sent to an unapproved model endpoint

Security·Investigating

Business impact

Possible data-protection violation

An ungoverned agent (marketing-gpt) is calling OpenAI without an owner

Platform·Open

Business impact

No accountability or spend control

Egress is default-allow org-wide

Security·Open

Business impact

Any agent can reach any endpoint

At A Glance

Governed agents

32 / 38

6 agents not yet under policy.

Egress stance

default-allow

Org-wide; not yet inverted.

Sensor coverage

2 gaps

Two environments unmonitored.

Spend

$42.8k / $50k

86% of the monthly budget.

For The Board

Governance improved four points this week to 72, but remains below our target of 80 and still reads as needs attention. The single biggest exposure is egress: the org runs default-allow, so any agent can reach any endpoint. Three high-risk findings are open, including customer PII sent to an unapproved model and an ungoverned agent calling OpenAI with no owner. Coverage and identity are healthy; egress control and spend discipline are the focus areas for next quarter.

Last exported 6 days ago