Agentic AI security moves from theory to budget line
Agentic AI security is the set of tools, policies, and real‑time control layers that keep autonomous, AI‑driven software agents from abusing permissions, mishandling data, or executing workflows that violate enterprise risk and compliance standards, even as those agents operate independently across applications, endpoints, and identities. Neo has emerged from stealth with USD 100 million (approx. RM460 million) in funding to build security and control infrastructure for agentic software. Glow exited stealth with USD 180 million (approx. RM828 million) to advance a prevention‑first approach to endpoint security, and Empirical Security raised USD 25 million (approx. RM115 million) in Series A funding to expand its predictive cybersecurity platform for exposure and vulnerability management. That combined USD 305 million (approx. RM1.4 billion) is not venture hype; it is a blunt admission that AI agents are arriving faster than enterprises can govern them.
Neo: Building an operating system for autonomous software control
Neo’s thesis is straightforward and unsettling: traditional security was built for predictable applications and human users, not for fleets of AI agents acting on valid permissions while mimicking user behavior. Neo describes itself as an Agentic Software Control company, promising visibility and policy controls across AI agents, AI‑enabled applications, browsers, identities, plugins, extensions and conventional software gaining autonomous capabilities. Its real‑time control layer watches how agentic software behaves, helps companies identify which tools are operating across their environments, what they can access, and enforces policies before risky actions occur. The platform ties each action to the responsible human, agent, application or identity, creating a much‑needed audit trail when multi‑agent workflows blur accountability. Organizations can set policies for tool calls, API access, data movement and agentic workflows, with native enforcement that can block risky activity and malicious models directly. This is AI agent governance in practice — not more alerts, but a brake pedal built into the software stack.
Glow: Prevention‑first endpoint security for the AI era
Glow’s founders treat the endpoint as the front door for enterprise AI — and right now, that door is wide open. AI has changed the endpoint into the entry point for enterprise AI as employees adopt new tools, connect agents, and integrate AI into workflows faster than security teams can review. Regular AI use on corporate devices, whether authorized or not, has risen from 15% to 45% in one year. Glow responds with a prevention‑first platform: it gives security teams control over everything that runs on the endpoint, using specialized AI agents that continuously map the environment, analyze risk in real time, and enforce policies automatically. These agents proactively decide which software is allowed in and which should be removed, without slowing down the business. Its context and reasoning engine makes prevention adaptive in a way that was not possible before, shrinking the endpoint attack surface while still letting employees adopt and use AI securely. In other words, Glow argues that in enterprise AI safety, "react and remediate" is obsolete; "predict and prevent" is the only model that scales.
Empirical Security: Predictive exposure management in a machine‑speed threat world
If Neo and Glow confront autonomous software control in real time, Empirical Security tackles the quieter crisis: vulnerability lists that grow faster than security teams can think. The company has raised USD 25 million (approx. RM115 million) in Series A funding to expand its predictive cybersecurity platform for exposure and vulnerability management. Its Foundation model monitors more than 18,000 exploited Common Vulnerabilities and Exposures to highlight active threats versus issues that pose less immediate risk. Radiant, a customized predictive engine, blends threat intelligence with each customer’s environment, risk thresholds and security priorities to identify the exposures most relevant to that business. This shifts vulnerability work from generic scores to localized, evidence‑based risk forecasts, helping security teams prioritize remediation, justify decisions to executives and avoid spending scarce time on low‑impact issues. Empirical expects demand to grow as artificial intelligence lets attackers find and exploit weaknesses faster. In that context, predictive exposure management is not a nice‑to‑have; it is table stakes for enterprise AI safety in technology, healthcare and financial services, where attack surfaces and stakes are both enormous.
The new AI security stack: governance, prevention and prediction
Taken together, Neo, Glow and Empirical Security show what the next AI security stack will look like: agentic AI security wrapped around every point where autonomy meets enterprise data. Neo puts AI agent governance at the control plane, watching agentic workflows, enforcing fine‑grained policies and providing native controls when autonomous software crosses red lines. Glow turns endpoints into governed gateways, assuming attackers now enjoy "Mythos‑class" capabilities and treating every tolerated risk as exploitable at machine speed. Empirical Security brings predictive exposure intelligence to overwhelmed teams, narrowing thousands of vulnerabilities into the few that matter most. The practical impact for ordinary users is clear: fewer arbitrary blocks and surprises, and more intentional guardrails that keep AI helpful without letting it run wild. Enterprise leaders who keep deploying agents without these kinds of safety guardrails are not being bold; they are being reckless. AI needs freedom to automate, but that freedom now has a condition: trustworthy, prevention‑first control over autonomous software behavior.
