Veeam argues AI agents force security and recovery into one discipline

In Spain, a company recently reported to its regulator that a mostly autonomous AI agent had broken in through a weakness in a running application and then altered personal data and billing records ^[17]. Veeam chief executive Anand Eswaran opened the launch with that case and treated the 72-hour disclosure requirement as an engineering problem ^[7]. "Somebody has to answer three questions: Which records did it change? What did it change them to? And what were they before?" he said ^[8].

SiliconANGLE's analysis separates modification from the incidents security teams already plan for. Ransomware encrypts and exfiltration steals, and both are detectable events with established playbooks ^[10]. Modification is harder. "You can disclose a theft. You cannot disclose your way out of a modification," Eswaran said. "You have to reverse it, and you can only reverse what you can identify precisely." ^[9]

"A recovery point tells you what you can restore," Eswaran said. "A recovery map tells you what to restore first, what changed, and what not to bring back." ^[11] Each of the three questions is about what changed, the part Eswaran assigns to the map.

Eswaran also pointed to the agents nobody chose. "You can govern the agents you choose. You cannot register the ones that you do not know about," he said ^[13]. By the survey's count, more than 82% of organizations have unknown agents. Veeam argues that the data is the one vantage point that does not require knowing the actor in advance, because every agent eventually touches it ^[16]. It is also an argument for a product that sits on the data, so it should be tested before it is bought.

The survey figures come from Veeam's own report ^[12]. The account of the launch does not include a price for the add-on or any customer results.

A two-by-two settles whether the budgets need to merge for a given data store. One axis is whether you can list every agent that touches the store. The other is whether you can reverse a change to a single record. Where you can do both, the budgets can stay apart. Where you can list the agents but cannot reverse, you have governance and no answer to "what were they before?" Where you can reverse but cannot list, you can recover changes you cannot yet explain. Where you can do neither, all three of Eswaran's questions go unanswered inside the 72 hours.

We'd start with the stores in that last box and spend on reversal before inventory, since the 72-hour requirement does not wait for an inventory. The tradeoff is scope. So far iResOps covers only Microsoft 365 ^[1], so every other store stays with the tools you already have, and a shared budget needs one named owner.