Google Cloud IAM Unified Access Policies Now in Preview: One Policy Can Bind Up to 25 Targets at No Extra Charge
Google Cloud has announced that Unified Access Policies, the next-generation access control engine for Cloud IAM, is now in preview. It combines allow and deny rules and treats agents as first-class principals; preview use carries no extra charge.
News date: About 6 min read
Contents
Google Cloud announced on its official blog that Unified Access Policies, the next-generation access control engine for Cloud IAM, is currently in preview rather than general availability, and Google says there is no extra charge. Cloud IAM is the identity and access management service that decides who can do what on which cloud resources. UAP lets administrators write allow and deny rules in a single policy and bind it to multiple resources at once. For platform and security teams managing many projects and AI agent identities, this may reduce the burden of maintaining duplicate policies.
How do Unified Access Policies differ from existing IAM policies?
According to Google Cloud, the traditional model has three problems. First, allow and deny use different syntax. Second, policies are attached directly to individual resources and maintained one by one as "snowflakes." Third, a single allow policy has a 64 KB limit, which large enterprises routinely hit as agent identities grow. As an example, Google says that applying a standard baseline to 25 projects would require maintaining 50 allow and deny policy configurations in Terraform (a tool for describing and managing cloud infrastructure as code).
UAP merges allow and deny rules into a single schema and separates policy definition from resource enforcement. Google says this means a policy is written once and then bound across the whole organization. UAP applies to principals (the identities being granted access) such as people, workloads and agents. At enforcement time, Cloud IAM evaluates existing allow policies, deny policies and UAP policies together.
| Aspect | Existing approach (as described by Google) | UAP preview (as described by Google) |
|---|---|---|
| Rule syntax | Allow and deny use two different syntaxes | Single schema; one policy can contain both allow and deny |
| How applied | Policy attached directly to individual resources | Bound independently; one policy can bind to up to 25 CRM targets |
| Scale | 64 KB limit per allow policy | Up to 5 policies per resource, each up to 64 KB and 100 rules |
| Principal capacity | Not described by Google | Up to 2,500 principals and 500 groups per resource |
| Agent identities | Not described by Google | Agents are first-class principals; rules can target individual agents or agent pools |
| Interfaces | — | Console, gcloud CLI, Terraform and Cloud IAM v3 REST API all support CRUD |
What features does the preview offer, and who benefits?
- Separation of duties: Google says central security architects can write baseline and guardrail policies at the organization node. Guardrails are firm rules that block risky actions. Folder and project administrators can then bind these policies themselves, but local teams cannot modify the central rule logic.
- Permission wildcards: prefix, suffix and substring wildcards can be combined with predefined roles and excluded permissions, replacing permission lists that tend to break.
- Agent readiness: UAP handles agent principals in the principal://agents.global... form and can work with short-lived, narrowly scoped tokens. Google also describes it as a foundation for human-in-the-loop setups (where a person stays involved in agent decisions) and delegated agent authorization.
How do you create and bind your first policy in the console?
Google Cloud explains that this can be done in three steps in the console, or automated with Terraform, the gcloud CLI or the REST API.
- In the resource selector, choose an organization, folder or project, go to IAM & Admin > IAM, open the new Unified Access Policies tab, and click + Create policy.
- In Policy Details, enter a name and description. Then, under Add rules, choose the principals, the rule effect (Allow or Deny) and the permissions. You can use + Add exception principal to exempt a breakglass (emergency access) administrator group. To combine allow and deny rules in the same policy, click Add another rule.
- In the Enforcement step, choose the organization, folder or project scope where the policy should take effect, and click Bind policy. Afterward, you can review the rules and bindings in Policy details, or click View JSON to export them for IaC (infrastructure as code) workflows.
What limits should you know before adopting it?
The scale figures Google gives fall into two kinds. Each CRM resource can have up to 5 policies, 2,500 principals and 500 groups; each policy can be up to 64 KB with up to 100 rules. All of these figures are Google's description of the preview. In addition, UAP is evaluated alongside existing policies, so your current allow and deny policies still take effect at enforcement time. For detailed configuration, see Google Cloud's official UAP documentation.
Frequently asked questions
Is Unified Access Policies generally available now?
No. Google Cloud says UAP is currently in preview and states that the preview carries no extra charge. Google has not given a general availability date.
Will UAP replace existing IAM allow and deny policies?
Google does not mention replacement. Google explains that at enforcement time Cloud IAM evaluates existing allow policies, deny policies and UAP policies together.
Which tools can manage UAP?
Google Cloud says UAP offers full create, read, update and delete functionality in the Google Cloud console, the gcloud CLI, Terraform and the Cloud IAM v3 REST API.
What is different about UAP for AI agents?
According to Google, agents are treated as first-class principals and do not need to share credentials with ordinary service accounts. Administrators can set allow and deny rules for individual agents or agent pools, and can pair them with short-lived, narrowly scoped tokens.
Where can I view and export a policy's contents?
On the Unified Access Policies page in the console, open Policy details for a policy to audit its rules and bindings. Then click View JSON to export it for IaC workflows.
Found this guide helpful? Share it with someone who needs it
Cloudflare WAF Blocks Ruby ruby2shell (CVE-2026-132825) by Default on All Plans, Including Free; Upgrading to 3.3.14, 3.4.13 or 4.0.10 Is Still Required
Cloudflare says it has added two default-Block WAF rules protecting against the critical Ruby vulnerability ruby2shell, covering Free through Enterprise plans; it stresses the WAF is only a compensating control and Ruby must still be upgraded.
PHP2Shell: Two Critical PHP Flaws (CVSS 9.2) Are Patched, but Dell Says You Should Also Rotate Credentials and Validate Backups
NVIDIA and Microsoft at Ignite: Vera Rubin NVL72 in Preview on Azure, Nemotron 3.5 in Foundry, Windows 365 Direct Stream in Limited Preview
Google Quantum AI Makes the Cover of Science: Quantum Contextuality Holds at Up to 105 Qubits, With a 98.3% Win Rate in a Coordination Game
Sources
Compiled from the references and verification dates listed above; fares, schedules, product features, and terms can change, so please confirm the latest details on the relevant official website before travelling, booking, or purchasing.