HP argues that secure adoption of artificial intelligence in business starts on the device itself, not in the cloud, and proposes visibility and isolation as the first lines of defense against AI agents. That’s the view of James Wright, HP’s global head of personal systems security solutions, in an interview with CRN that the company reproduced on its press blog on November 2, 2026. It’s sponsored content and a vendor’s perspective, not an independent study.
AI security at the endpoint according to HP, in 20 seconds
- HP argues that AI is moving from the cloud to the device, and that security needs to follow.
- Wright frames visibility into which AI tools are in use as problem number one.
- He proposes applying to AI agents the same isolation HP already uses against malware.
The piece builds on an idea that has circulated in the PC industry since the arrival of machines with neural processing units (NPUs): more and more AI workloads will run on the employee’s laptop or desktop, not in a remote data center. HP stresses that processing AI locally can help keep sensitive data and intellectual property inside the organization, besides improving performance and user experience. But it adds a condition: those advantages only materialize if there’s visibility, protection, and governance at the endpoint itself.
Visibility, HP’s problem number one
“Problem number one is visibility,” Wright says in the interview. As he frames it, organizations need to know which AI tools are running on their machines, what data and systems they can reach, and where unauthorized use is showing up — what the industry calls shadow AI. Without that inventory, HP argues, it’s hard to set limits or make informed adoption decisions.
The company proposes that visibility be built into the adoption process itself rather than bolted on afterward. Decisions about access, oversight, and accountability, it argues, should move in lockstep with decisions about where AI creates value. It’s a reasonable framing, but the piece doesn’t specify which HP tools deliver that visibility or what data they collect.
Isolation against the risks of AI agents
The second pillar is agentic AI. HP warns that autonomous agents — which access data, interact with systems, and take actions within workflows — introduce new risks. They also hand attackers more automation and speed, which the company says could turn previously low-risk areas into attractive targets. Its recommendation is to govern what agents are allowed to do and access, and to build security into the AI architecture from the start. Security vendors have been racing to add exactly that kind of control, from network-level kill switches that can cut off risky AI agents to endpoint tools built specifically to rein in what agents can touch.
To get there, Wright points to isolation as a capability HP already has. He explains that the company uses containment technology in several of its security products to keep potentially malicious activity confined to secure environments, and that the same approach could be relevant to some of the challenges of agentic AI because it helps shrink the attack surface. The blog doesn’t name specific products, but that technology underpins HP Wolf Security and Sure Click, which open attachments, web pages, and downloads from untrusted sources inside micro virtual machines isolated from the rest of the system.
That same isolation layer feeds the company’s threat reports. In the latest one, published in March 2026 and covering the fourth quarter of 2025, HP reported that at least 14% of the email threats detected by Sure Click had made it past one or more email gateway filters, and that attackers were mainly using AI to produce malware faster and with less effort. The company says its customers have opened more than 60 billion attachments, pages, and downloads with Wolf Security without a reported breach — a figure that comes from HP itself. This outlet has previously covered another of its warnings, about the rise of attacks that abuse legitimate remote-access tools as a backdoor.
What the piece doesn’t clarify is how that isolation would apply to an agent that, by definition, needs access to real data and systems to be useful. Wright says the approach “could be relevant” to “some” of the challenges — a cautious phrasing that announces no specific product or timeline.
Hardware, software, and the channel’s role
HP points to making both the hardware and part of the software of its machines as an advantage, which, according to Wright, gives it more options for designing protection in from the device up. Other PC and chip makers make similar arguments: Intel and CrowdStrike, for instance, have already laid out their own approach to strengthening AI PC security against this new local-device risk. Other vendors are tackling the device side of the same problem too — Commvault, for one, now protects the work AI agents leave behind on employee laptops.
The last part of the piece is aimed at HP’s resellers and partners, consistent with its original publication in CRN, a channel-focused trade outlet. The company argues that its understanding of each customer’s infrastructure and needs can turn uncertainty into concrete decisions about protection and deployment. Wright closes with a line that sums up the message: “We need greater governance and control to give businesses the confidence to embrace it” — in other words, more governance and control are needed before businesses will feel confident adopting AI.
The blog post lands in the middle of a leadership transition at the company, which just hired Emma Walmsley as its new chief executive. A footnote points to HP’s Work Relationship Index, a survey of 19,506 office workers across 15 countries conducted with National Research Group between April 7 and May 13, 2026, although the body of the text doesn’t cite any data from that study.
Frequently asked questions
Why does HP say AI security starts at the endpoint?
Because, in the company’s view, more and more AI workloads run on the device itself, and that requires visibility, protection, and governance there to capture benefits such as keeping sensitive data inside the company.
What is the isolation HP proposes for AI agents?
It’s the technique HP already uses in products like Wolf Security to open risky content inside environments separated from the rest of the system. HP says it could help shrink the attack surface for agents, but it hasn’t announced a specific product for that purpose.
Who is James Wright?
He’s HP’s global head of personal systems security solutions. His comments come from a sponsored interview published by CRN and reproduced on HP’s blog on November 2, 2026.
Sources:
- HP, Securing AI At The Edge (11/02/2026)
- CRN, sponsored interview with James Wright
- HP Wolf Security Threat Insights Report, Q4 2025 (March 2026)
This article is also available in Spanish: “HP sitúa el endpoint como primera línea para adoptar la IA con seguridad“, on revistacloud.com.
