Zero Networks launches AI segmentation to curb shadow AI

Zero Networks has launched AI Segmentation for its security platform, aimed at giving organisations direct control over AI agents.

The new capabilities target three areas many security teams are now trying to manage at once: access by AI tools and agents, the risk of AI-driven lateral movement inside corporate networks, and the growing burden of compliance and risk work. Zero Networks says its approach uses network identity controls to govern what AI systems can reach and to limit how activity can spread across an environment.

The launch reflects a broader shift in cyber security as businesses move from simply identifying AI use to setting rules around it. Many organisations are now grappling with a mix of sanctioned AI tools, unsanctioned services accessed by staff, and emerging autonomous agents that can interact with internal systems with limited oversight.

The platform can govern access to cloud AI services such as ChatGPT, Gemini and Copilot at the network layer, allowing companies to permit approved services while blocking unauthorised shadow AI tools.

Another part of the release focuses on AI agents operating within enterprise environments. Customers can identify which agents are running, what resources they are accessing and how they communicate, while applying the same identity-based controls used for people and devices. This allows organisations to set least-privilege boundaries around each interaction rather than relying only on visibility or post-incident investigation.

The platform also includes controls for large language model infrastructure. Zero Networks says this is intended to ensure that only authorised systems can connect to model environments, reducing the risk that compromised systems or manipulated inputs reach sensitive resources.

Containment focus

A central theme of the product update is lateral movement, a long-standing cyber security concern that refers to how attackers move across networks after gaining an initial foothold. Zero Networks argues that AI can increase the speed and scale of that movement, either through hostile use of AI or through poorly governed autonomous agents making connections that security teams did not intend.

To address this, the platform removes unnecessary connectivity across the environment and enforces least-privilege access based on identity and network policy. The goal is to stop unauthorised users, systems or AI agents from reaching critical systems regardless of the point of entry.

This matters because many current security programmes still focus heavily on endpoint detection, access monitoring and dashboards. Network segmentation, by contrast, aims to reduce the pathways available in the first place. Vendors in this part of the market argue that containing movement after an intrusion can be as important as preventing the initial breach.

Compliance layer

Zero Networks has also added an AI-powered compliance and risk engine to the platform. Security teams can use natural language queries to examine live network activity across large volumes of connections, while the system evaluates activity against frameworks including NIS2 and CIS Benchmarks.

The feature is intended to help organisations under pressure to demonstrate continuous compliance rather than rely on periodic manual reviews. Dynamic risk scoring and gap identification are now common themes in security products as firms try to reduce the effort involved in linking technical controls to regulatory requirements.

"Most vendors are out there selling AI hype. We're not. Zero Networks puts enterprises in control of AI - full stop," said Benny Lakunishok, Chief Executive Officer and Co-Founder, Zero Networks.

He added: "This isn't just visibility. It's real control. Real-time, deterministic control over AI agents, combined with AI-driven visibility and an integrated Compliance and Risk Engine that continuously scores risk, maps activity to frameworks like NIS2 and CIS, and flags what actually matters. While others are still watching dashboards, Zero is enforcing outcomes - stopping lateral movement and preventing threats from becoming business problems."

Zero Networks is positioning the release as part of a broader argument that AI security should be handled through enforceable policy rather than observation alone. That message comes as security teams face pressure to manage both employees' use of public AI services and the rise of internal agents with machine-driven decision-making.

The new capabilities are available now as part of the existing Zero Networks platform. The company's broader product line centres on zero trust security and microsegmentation, with an emphasis on reducing unnecessary network access and limiting the impact of breaches once attackers or malicious software gain entry.

For buyers, the key question is whether tools marketed for AI governance can move beyond inventory and alerts to practical control over traffic, identity and access. Zero Networks is arguing that network-level enforcement is where that control should sit.