The FusionAuth team is at Google Cloud Next. Each year, Google Cloud serves as a litmus test for where much of the technology industry is headed. This year the focus is on governance, and the keynote held a few gems that are worth unpacking.
The New Reality: Managing Chaos#
This year has an interesting shift in conversation. It has gone from "can we build an AI agent" to "how do we manage thousands of them."
Google's answer is the Gemini Enterprise Agent Platform. It's a full stack for building, deploying, and governing agents across an organization. Agent identity is central to this stack. Every agent gets a unique cryptographic ID, defined authorization policies, and a traceable audit trail. Google calls this "governance at machine speed." They are promising that you'll know what every agent did, be able to prove it, and revoke its access when needed.
Then came the 'Wiz Show'. Google completed the acquisition of the cloud security firm in March of this year for $32 billion. The keynote put a spotlight on why it happened: shadow AI. The term is newly-coined, but the problem is something we're all familiar with — unauthorized models and agents running inside enterprises, outside of anyone's control. This is the threat that Google is most worried about. Wiz's new AI Application Protection Platform promises visibility across the AI stack so you can find agents you didn't know existed, map their access, and validate the risks before attackers do.
Wiz's demo was compelling — if all you need is an alert after a breach. What the findings showed made our ears perk up: an unguarded AI agent with a direct line to a sensitive database.
$32 Billion Questions#
It's an unfortunate truth that posture management often starts after the agent already has the keys. An agent calls your app's API and that's when the questions begin.
- Who issued the credentials?
- Who decided what the agent was allowed to do?
- Who revokes the credentials when something changes?
Posture management tells you what's exposed. Orchestration controls what runs. Having these priorities reversed is a bit like telling someone your front door was open after you find that all of your furniture is missing.
Who issues the token, what it permits, and on whose behalf are all still identity questions. The effect lands squarely on every application those agents are calling.
Google as a Weather Vane#
It's hard to miss the point when Sundar Pichai, CEO of Alphabet, and Thomas Kurian, CEO of Google Cloud, both spend keynote time on agent governance. We're already in the heat of things with agentic enterprise activity. In fact, it's moving faster than the security infrastructure that surrounds it. That's why it's more critical than ever to prevent rather than simply detect.
Google Cloud COO Francis deSouza put a number to it on stage: the mean time to exploit a vulnerability is now down to minus seven days. Exploits are now routinely happening before patches exist. Finding a problem after an agent already has access isn't fast enough. The controls have to be in place from the start. Authentication, token issuance, and decisions around scoped access need to be made before the agent is running.
Wiz cofounder Yinon Costica put our new reality into plain language: someone in finance can vibe-code an agent in an afternoon. That agent connects to something, it acts on someone's behalf, and it has access that was granted deliberately — or it wasn't.
Google's answer is to find the agent, assess what it has done, and govern it within their stack.
That's a reasonable answer to the wrong question.
The more important question is why the agent had that access in the first place. A CTO explaining a breach to the board would rather describe what their system made impossible than what their platform eventually detected.
$32 billion buys you better visibility. The identity layer determines what there is to see.
