NetApp opens its storage console to the language models customers have already vetted

Jeff Baxter, NetApp's vice president of technical marketing, framed the release as an answer to what buyers already run. "Just about every enterprise we work with has, by this point, established one or more approved LLMs, often specific enterprise instantiations of them that they've validated, they've set up guardrails, and they've set up how they want to leverage that within their enterprise," he said ^[3]. He tied the update directly to the growing use of AI in data and infrastructure management ^[4]. The ChatOps feature accepts the model the customer selects, including domain-specific ones ^[2].

A company that has already validated a model and set guardrails on it reuses that work on its storage estate ^[2][3]. It avoids approving a second model that a vendor chose. Baxter called the Console "the front door to the NetApp platform," so the model a customer plugs in sits at the entry point to everything else it runs on NetApp ^[10].

The guardrails Baxter described were set for uses each enterprise chose before this feature existed ^[3]. Singh, who leads NetApp's Data Infrastructure Platform division, described the new use: an administrator speaks or types a request, and the system "will then be able to take that and convert that into the backend operations for storage infrastructure," he said ^[6]. Approving a model to answer questions is a different decision from approving it to provision storage and run remediation on live systems ^[5][7]. In my view, clearing a model for one does not automatically clear it for the other.

Each action now passes through two sets of controls. The customer's guardrails govern the model. NetApp's blog says its interface applies policy guardrails across ONTAP storage systems ^[5]. If an automated fix goes wrong, the post-incident review has to establish which layer should have stopped it. The ITPro report does not say how permissions are split between the two layers or how the Console records what a model did on a customer's behalf.

The case for results so far rests on Singh's account that initial testing delivered marked benefits for users ^[8]. He described adoption as a long process. "We're continuing to work very closely with customers as we go through the adoption curve, because it's always typically a journey," he said ^[11].

NetApp's blog describes autonomous storage operations as something the interface delivers now ^[5]. For customers already on ONTAP, the decision belongs to this quarter. It is narrow: which approved model to connect, and which operations it may run without a person signing off. Every remediation the model triggers ^[7] will sit in an audit trail shaped by two sets of rules, the customer's and NetApp's ^[5].