Veeam’s ‘Agent Commander’: Bringing Guardrails and Resilience to the Wild West of AI

For years, the tech industry has treated data backup and security as two distinct silos.

Backup was the boring “insurance policy” you hoped you’d never need, tucked away in the back office. Security was the “perimeter” designed to keep the bad guys out. But as ransomware began blurring those lines, a new category emerged: Data Resilience.

Now, we are entering a third act, fueled by the explosive — and sometimes reckless — adoption of generative AI and autonomous agents. Last year, backup and recovery leader Veeam acquired Securiti Inc for $1.7 billion to move into AI governance. While Securiti is great to scan backups for malware, the combination of Veeam and Securiti can make the company the “Command Center” for AI.

Recently, Veeam announced Agent Commander, the first major integration since the acquisition closed in December. This is the first step in evolving Veeam from being a security-adjacent backup and recovery company to a proactive AI resilience provider.

The ‘canyon’ problem: Why AI is stalled

During a recent briefing with Nikhil Girdhar, Veeam’s head of product marketing for Securiti, we discussed the “canyon” analogy, which I heard AWS CEO Matt Garman discuss at the recent Cisco AI Summit. If one needs to cross a deep canyon on a narrow board, you’re going to crawl because you’re terrified of falling. But if you put up sturdy guardrails, you’ll run across.

Deloitte has reported that 74% of enterprises want to put AI agents into production, but only about 23% have actually done it. Why? Because they lack the guardrails. They are terrified that an autonomous agent — acting with the permissions of a human employee — will hallucinate, leak PII (Personally Identifiable Information), or accidentally delete a petabyte of production data.

To use Garman’s analogy, they are currently crawling. The IT leaders I have talked to are aware that if they can’t quickly move out of the crawl phase, they risk falling behind their peers who are running.

Enter Agent Commander: Detect, protect, undo

The Agent Commander announcement is built on a three-pronged framework designed to give CISOs the confidence to “run across the board.”

1. Detect AI (visibility into the shadow)

There’s an axiom in cyber that “you can’t secure what you can’t see,” and that extends to AI, as you also can’t govern what you can’t see. Agent Commander provides a Data Command Graph — a knowledge map that discovers every agent, copilot, and LLM interacting with your production and backup data.

• The Context Gap: Agent Commander is about more than just finding people who are using Microsoft Copilot. It’s seeing the interactions between agents, such as a custom-built agent in AWS, currently touching a SharePoint folder containing European residents’ data, potentially violating GDPR.
• Shadow AI: It identifies “unmanaged” agents that developers might have spun up for experimentation but left connected to live production databases.

2. Protect AI (foundational guardrails)

Protection happens at two layers: the data layer and the runtime layer.

• Data Control: Using the Securiti engine, Veeam can now automatically classify, label, and put controls around data across 300+ sources. If an agent sees a “HIPAA” label, it can be programmed to ignore that file.
• LLM Firewall: This is a crucial addition. It monitors the “prompts” going into an agent and the “responses” coming out. It can block prompt injection attacks, jailbreak attempts, or even prevent an agent from generating source code if it’s supposed to be a customer service bot.

3. Undo AI (precision resilience)

This is where Veeam brings some “magic” to a process that was historically slow and manually intensive. Traditionally, if an agent messed up a database, the entire volume would need to be restored. Veeam offers a faster and more automated process through the following:

• Surgical Recovery: Because Veeam now has file-level observability into agent activity, it can see that an agent “mistakenly” deleted exactly three files.
• The “Undo” Button: Instead of a full-scale restore, the admin hits “Undo” on those specific files and actions. It is precision resilience.

The business case: when Securiti pays for itself

One of the most compelling parts of the briefing was a case study to identify “ROT” data (Redundant, Obsolete, and Trivial).

By leveraging the platform’s visibility, they eliminated petabytes of useless data in less than a year. The result was millions in infrastructure savings. This changes the conversation for the IT admin. Usually, security is a cost center.

In this model, cleaning up your data “hygiene” to prepare for AI pays for the security tools themselves, something typically not seen in the security industry. Also, better data hygiene leads to:

1. Lower Storage Costs: Companies aren’t storing or backing up junk.
2. Lower Cyber Insurance Premiums: The attack surface is demonstrably smaller.
3. Higher AI Efficacy: LLMs aren’t hallucinating based on 10-year-old, outdated data.

The shift from back-office to boardroom

For Veeam, this shift from being “the backup guys” to a core AI resilience provider is important as it looks to IPO in the future.

The evolution of the company away from a product that’s associated with a legacy market, even though they are the market leader, to one that is a critical vendor in the AI era will drive much greater valuation. This is like Pure Storage’s repositioning away from storage, a market it dominates, to Everpure, which moves it into AI data management.

For the Veeam administrator, this is a massive career upgrade and is as important as jobs change because of AI. In Davos, the World Economic Forum released a report that stated AI would eliminate 92 million jobs but create 175 million by 2030. Many of the jobs eliminated will be those that revolve around mundane tasks, such as restoring backed-up files, where being responsible for safe AI will be a job in high demand.

By evolving its product and positioning, Veeam is enabling the people who use its product to be part of the 125 million new jobs and not the 90 million that will go away.

Final thought

In the world of Agentic AI, the speed of innovation is limited by the strength of your brakes. If you don’t trust your brakes, you won’t drive fast. With Agent Commander, Veeam is providing a high-performance braking system that allows the enterprise to finally hit the gas and run with AI safely.

Also read: Research shows AI agents are creating insider-style blind spots for security teams as enterprises race to govern non-human identities.

About Author

AndyC

Andy Curtis is an award-winning security consultant, researcher and public speaker. He has been working in the computer security industry since the early 1990s, having been employed by state and federal government, leading healthcare and banking providers across three continents. He has given talks about computer security for some of the world’s largest companies, worked with law enforcement agencies on investigations into hacking groups, and is a regular voice on TV and radio explaining IT security threats.

See author's posts