Skip to content

Google Cloud previews Unified Access Policies to combine allow and deny rules in IAM

Google Cloud has released Unified Access Policies in preview at no extra cost. The new engine merges IAM allow and deny rules into one policy schema that can be bound across up to 25 organization, folder, or project targets.

3 min readBased on 1 source1 views
CLOUD

TL;DR

  • Unified Access Policies (UAP) is in preview in Google Cloud IAM at no additional cost.
  • One policy can hold both allow and deny rules and be bound to up to 25 organization, folder, or project targets.
  • Each resource can carry up to five UAP policies, which raises the size and principal limits of legacy IAM.

What UAP changes

Google Cloud describes UAP as its next-generation access control engine inside Cloud IAM. The core idea is to separate policy definition from enforcement. An administrator writes a policy once and binds it to the parts of the resource hierarchy where it should apply, instead of attaching separate configurations to each resource. Google says UAP consolidates allow and deny rules into one schema, so teams no longer have to reconcile two sets of files and two syntaxes to express one access posture.

A policy binding is the link between a policy and a target node in Cloud Resource Manager (CRM), which covers organizations, folders, and projects. A single UAP policy can be bound across up to 25 CRM targets in preview, which removes the need for duplicate blocks in Terraform.

The problem it targets

The announcement frames the change around three pressures that build up in large cloud estates. The first is permission sprawl. Administrators either grant broad permissions through coarse roles or manage two different syntaxes to define a custom posture for one workload. The second is duplication. Attaching policies directly to resources means that a baseline applied across 25 projects can turn into 50 separate allow and deny configurations to maintain. The third is size. Google notes that large enterprises routinely reach the 64 KB limit on a single allow policy as non-human and agent identities multiply, which forces teams to refactor policies before they can onboard new teams.

Google also presents agents as a distinct identity type. UAP treats them as first-class principals, identified with a principal:// agents.global path, rather than as generic service accounts that share credentials.

Limits and how rules are evaluated

UAP lets teams attach up to five UAP policies to a single CRM resource. Each policy can be up to 64 KB and hold up to 100 rules. Google says this gives an effective ceiling of 2,500 unique principals and 500 groups per resource. The cover-to-cover numbers are the vendor's own, and they apply to the preview.

At runtime, Cloud IAM evaluates existing allow policies, deny policies, and UAP policies at the same time across the resource hierarchy. Google says the feature launches with create, read, update, and delete support in the console, gcloud CLI, Terraform, and the Cloud IAM v3 REST API. Wildcards are another part of the design: policies can use prefix, suffix, and substring wildcards, along with predefined roles and excluded permissions, in place of long explicit permission lists. Deny guardrails can be written with patterns such as compute.googleapis.com/firewalls.*.

For delegation, Google suggests a split in responsibilities. Central security architects write baseline and guardrail policies at the organization node, while folder and project administrators bind those policies locally without changing the central rule logic. Exceptions can exempt break-glass administrator groups.

Why it matters

For platform and security teams, UAP offers a way to manage one access rule set across many projects without copying it. Teams that currently maintain duplicated Terraform blocks for allow and deny rules could shrink that code, and they gain a way to keep deny guardrails in the same auditable policy as the grants they constrain.

For developers, the practical effect depends on how their organization adopts the preview. Application teams may see fewer policy refactors when they add services or agents, and agent builders get identities that can be scoped individually or by agent pool. Because the feature is in preview, teams should test it in a non-production project before relying on it, and check the current documentation for limits and supported surfaces.

Sources

This article was written by Blozhik from the reporting below. Read the originals for full detail.

GoHow to simplify cloud access with Unified Access Policies, now in previewGoogle Cloud Blog
Was this useful?

Keep reading

Discussion · 0

Sign in to join the discussion. Liking does not need an account.Sign inRegister
Loading comments…
blozhikDeveloper news from across the web, explained. Every story links to its original sources.