Safety & Security

Google says 2027 is the year AI agents reshape enterprise security

Google's Cloud CISO Perspectives forecast says AI agents will run security operations at scale by 2027 — and businesses that adopt them carelessly will widen their attack surface.

5 ways AI will change how you defend your business in 2027
5 ways AI will change how you defend your business in 2027geralt / pixabay
By Elena Vasquez6 min read

Why it matters

  • Google's security forecast for 2027, Cloud CISO Perspectives, predicts agentic AI will operate in security operations centers at scale.
  • Google Cloud's Ramya Chitrakar: "2027 is the year analysts stop chasing alerts and start shaping outcomes."
  • Google's Heather Adkins calls for zero trust to extend from the network layer to the transaction layer for AI agents.
  • Google calls agent sprawl the next iteration of shadow IT and urges specialized, orchestrated agent teams over generalist agents.
  • AI-driven threats can now probe, enter, and move laterally across a network in minutes, per the report.

Google expects agentic AI to influence cybersecurity at scale by 2027, when the alert queue as security analysts know it will quietly disappear. That is the core claim of the company's new security forecast, Cloud CISO Perspectives, which argues that AI has already outpaced human-only response teams — and that businesses adopting AI security tools without the right precautions could grow their attack surface instead of shrinking it.

The shift is already underway. At the beginning of the year, few organizations used large language models and AI agents in their cybersecurity workflows, including for discovering and patching vulnerabilities. According to Google, that is now happening at scale. Speed and scale are the two main reasons enterprises consider AI for security operations centers (SOCs) and defense more broadly, and they anchor the forecast.

Here are the five ways Google says AI will change how businesses defend themselves in 2027.

What happens to the alert queue?

Alert fatigue is a well-documented problem in SOC teams. Analysts trawl alert after alert, many of them irrelevant or false positives, tying up skilled staff in tedious, time-consuming triage work.

The math has changed. Autonomous, AI-driven threats can now probe a network, gain entry, and move laterally across it in minutes. Chasing threat data and security alerts logged days or weeks ago across multiple systems cannot mitigate attacks moving at that speed.

Google's forecast says AI will mean the alert queue quietly disappears. Ramya Chitrakar, VP of engineering at Google Cloud Security, puts it directly: "2027 is the year analysts stop chasing alerts and start shaping outcomes."

The mechanism is an inversion of the traditional alert model. Instead of human defenders triaging alerts at the start of a cybersecurity incident, the alert becomes the final output, with autonomous agents handling "data enrichment, correlation, and triage automatically, surfacing a verified conclusion rather than a raw ticket," according to Google. Human cybersecurity experts would oversee alert handling and verify automated actions, freeing them for other tasks.

Why do AI agents need identities?

Custom agentic AI is often designed — or at least intended — to work autonomously, with human oversight. Thousands of individual agents working across multiple environments create high-value entry points into corporate networks, and cyberattackers target them.

Google's answer: treat each agent less like a tool and more like a human employee. Every agent needs its own identity to operate within a zero-trust architecture, retain control, and reduce the risk of compromise.

Heather Adkins, VP of security engineering at Google, thinks this needs to go further. She suggests the zero-trust concept should not stop at the network layer — it needs to reach the transaction layer.

That means establishing a paper trail for each agent, which is impossible to handle manually. Even within a zero-trust framework, AI agents could be given too much scope. Google recommends verification for each transaction and every authorization, plus a least-privilege set of credentials for each AI agent that can be revoked if there is a security incident or the agent is no longer required.

"When backed by real-time autonomous reasoning, we can dynamically allow a transaction on Tuesday and block the exact same request on Wednesday if conditions look anomalous," the report notes.

How does faster software development change security?

AI is accelerating the development and deployment of new software, and Google says security must become a continuous system embedded in the software delivery pipeline by 2027.

The threat backdrop is familiar. Threat actors have carried out software supply chain attacks for years, with the SolarWinds incident of 2020 among the most notable. Sandra Joyce, VP of Google Threat Intelligence, says cybercriminals are now poisoning everything from open-source code to software libraries — which AI coding assistants often pull automatically into codebases.

The old model is broken, in Google's view. Bug-fix tickets that take days to respond to, and security work crammed in at the end of a development cycle, no longer function. Writing and securing code must happen simultaneously, with the help of agentic AI, so speed doesn't compromise security.

"It's also crucial that we link live production environments directly back to source code so that defenders can trace a real-world exposure to the exact line of code, validate whether it is genuinely exploitable, and deliver a pre-tested fix at machine speed without breaking production," Google added.

Why are red teams targeting AI now?

As organizations rush to implement AI in their applications and services to appear "AI-first," innovative, and modern, security may be ignored or treated as an afterthought. The result, according to the report, is a new and broader attack surface than many businesses realize.

AI introduces its own security issues, including:

The attacker side has democratized too. AI is now in the hands of all cyberattackers, from script kiddies to state-sponsored groups, which makes auditing corporate networks and infrastructure crucial.

Penetration testing and red team exercises now need to cover agentic AI — a scope that many red teams didn't have only a few years ago. Google says organizations need to identify and fix security issues introduced by AI before threat actors take advantage of them.

The recommended workflow: CISOs should task pen testers with auditing before an autonomous workflow hits production, then combine the results with threat intelligence so defenders have what they need to bolster network defenses and harden classifiers against new and novel attacks.

What is agent sprawl — and why does Google compare it to shadow IT?

The fifth trend for 2027 is tackling agent sprawl by building coordinated teams of AI agents that can rapidly handle security tasks at scale.

Single AI agents given general instructions won't cut it. AI-assisted attackers can compromise endpoint devices, enter networks, move laterally, or start stealing data in a far shorter timeframe than before. Google says CISOs should replace these "generalist" agents with specialized, orchestrated teams that can meet the demands of defending against AI-driven attacks using specialist tooling and skill sets.

Each agent needs its own:

  • domain training and understanding,
  • memory,
  • dedicated tools,
  • the ability to act on its own within set limits.

A central coordinator needs to know when to hand a decision to a human. That requires oversight. In return, Google argues, organizations avoid agent sprawl — which the company believes is the next iteration of shadow IT and its associated threats.

Why this matters

The stakes cut both ways. The same agentic capabilities that let defenders triage at machine speed also arm attackers who can traverse a network in minutes, and Google's forecast treats that asymmetry as settled reality rather than prediction. Businesses that give agents identities, least-privilege credentials, transaction-level verification, and red-team coverage before production can shrink their exposure. Those that don't will discover that agent sprawl carries the same price tag shadow IT always has — only at machine speed.

Original: cloud.google.com

Share this article:

More from Elena Vasquez

Elena Vasquez

Show full bio

Market editor covering technology and product news in artificial intelligence and machine learning at AI In Context.

708 articles

Related articles

  1. Google Cloud warns attackers will turn from data theft to AI decision manipulation
  2. Google Reports First Autonomous AI Attack That Finds and Exploits Flaws
  3. BeyondTrust Predicts First Autonomous AI Breach by 2027
  4. SAS: Corporate AI's 2027 Test Will Be Trust, Not Deployment Speed
  5. Anthropic Warns Small AI Models Could Self-Replicate Across Devices

« Previous articleNext article »