Former SentinelOne executives launch Neo Security to secure AI agents across enterprise systems

Neo Security, Inc. emerged from stealth on July 20 with $100 million in combined seed and Series A funding from Andreessen Horowitz, Bessemer Venture Partners, Craft Ventures and Merlin Ventures. The Boston-based cybersecurity company is building software designed to discover AI-enabled applications, identify the permissions and data available to them, and block or hold actions before sensitive systems are affected. Neo Security was founded by former SentinelOne executives Nick Warner and Shlomi Salem alongside EasySend co-founder Eran Shirazi, giving the company an experienced commercial and technical leadership team. The funding arrives as ordinary enterprise applications, browsers and developer tools gain the ability to invoke models, call external tools and perform multi-step workflows with limited human supervision. The central tension is whether Neo Security can establish a new control category around agentic software before larger endpoint, identity and cloud-security providers absorb the same capabilities into existing platforms.

Why does Neo believe AI-enabled enterprise software requires a new security control layer?

Most enterprise security systems were designed around relatively predictable actors. A human user signs into an application, an endpoint runs approved software and an identity platform determines which resources that person can access. Security teams can monitor the device, account, application and network connection because the expected behaviour is broadly defined.

Agentic software changes that model by allowing applications to reason through tasks, choose tools and take actions using delegated permissions. A sales application may export customer records, a coding assistant may alter a production repository and a finance agent may update accounting systems. Those actions can be authorised at the identity level while still creating operational or data-security risk.

The immediate problem is not limited to specialist AI agents introduced through a formal technology programme. Existing software can acquire agentic functionality through an ordinary update, extension or plug-in, meaning an application trusted for several years may begin taking actions that were not anticipated when access was originally granted.

Neo Security is targeting the resulting visibility gap. The company argues that enterprises need to know which agents, models, browser extensions, Model Context Protocol servers and AI-enabled applications are present, what they can access and how they behave when running. Gartner expects task-specific AI agents to appear in 40% of enterprise applications by the end of 2026, compared with less than 5% in 2025, increasing the urgency of that inventory problem.

The strategic opportunity is therefore broader than protecting a chatbot. Neo Security is positioning itself around the control of software that can act with legitimate credentials, which makes the threat harder to classify than conventional malware. An action can be technically authorised, commercially useful and still inappropriate because the agent received excessive permissions or misunderstood the requested task.

What does Neo’s platform control that conventional endpoint security may not see?

Neo Security says its platform creates a live inventory of conventional software binaries alongside agents, scripts, skills, extensions and Model Context Protocol servers. It then examines how those components are configured, which identities and models are involved, what data they can access and whether permissions are broader than the intended business purpose.

The company’s proposed differentiation lies in runtime attribution and enforcement. Rather than recording only that a user account accessed a file or application, Neo Security is designed to connect the human identity, delegated agent and resulting action. That distinction matters when an employee authorises an agent to perform a task but the agent takes additional steps that the employee did not explicitly review.

Neo Security says policies can allow, block or temporarily hold actions for approval before they reach sensitive systems. An attempted export of a large customer database could be paused, access to cloud credentials could be blocked and an authorised code change could be permitted while creating an audit trail.

That model could complement endpoint detection, identity management, data-loss prevention and security-information platforms. Endpoint products generally focus on malicious code, processes and device activity, while identity systems determine whether an account has access. Neo Security is attempting to evaluate whether the action taken through that access is appropriate for the agent, context and data involved.

The platform also performs analysis on the device, according to the company, reducing the need to send sensitive information into an external inspection service. That could improve its appeal in financial services, healthcare, energy and other regulated sectors, although customers will need independent evidence that local processing, policy enforcement and endpoint performance operate as claimed.

The architecture is commercially sensible because the endpoint is where browsers, coding tools, business applications and agents often meet. However, it also places Neo Security in a technically demanding position. The platform must observe enough activity to understand agent behaviour without slowing applications, creating compatibility problems or becoming another highly privileged component that attackers attempt to compromise.

Why did investors commit $100 million before Neo disclosed revenue or valuation?

The launch announcement did not disclose Neo Security’s valuation, current revenue, customer count or the amount allocated between the seed and Series A rounds. It said the platform had been tested with organisations in financial services, transportation and energy, but did not identify those users or state whether they were paying customers.

The investment therefore reflects substantial confidence in the team and market category rather than publicly demonstrated commercial scale. Andreessen Horowitz and Bessemer Venture Partners are effectively funding the company to establish product maturity, expand engineering and build a go-to-market organisation before the agent-security market becomes crowded.

A $100 million launch provides advantages that smaller cybersecurity startups may struggle to match. Neo Security can recruit experienced detection engineers, support enterprise proof-of-concept deployments and invest in product integrations without immediately optimising for short-term revenue. It can also fund a longer sales cycle, which is important because large companies rarely deploy endpoint-level security controls without technical testing, privacy review and procurement scrutiny.

The size of the funding also raises the commercial threshold. Neo Security must eventually generate enough recurring revenue and customer expansion to justify a capital base more commonly associated with a later-stage security company. A technically interesting product will not be sufficient if enterprises treat agent security as an experimental budget rather than a permanent control category.

The absence of a disclosed valuation makes it impossible to judge how much ownership investors received or what future exit value is already embedded in the financing. That restraint is appropriate. The strategic question is not whether the funding creates a theoretical unicorn, but whether it gives Neo Security enough time to convert early interest into repeatable enterprise deployments.

How does the founders’ SentinelOne experience strengthen execution without removing startup risk?

Nick Warner previously served as Chief Revenue Officer, Chief Operating Officer and President at SentinelOne, helping build its commercial organisation and supporting the company through its public listing. SentinelOne said Warner joined when the business was still at an early revenue stage and later led its broader go-to-market operations.

That background matters because cybersecurity startups often fail through commercial execution rather than product weakness. Enterprise security sales require credibility with chief information security officers, strong channel relationships, technical support and the ability to prove that a new platform can coexist with established controls.

Shlomi Salem brings detection engineering and threat-research experience from SentinelOne, while Eran Shirazi previously co-founded EasySend and worked on enterprise software development. Neo Security has also recruited personnel with experience at Wiz, Palo Alto Networks and Israel’s Unit 8200 cyber and intelligence organisation.

The combined experience should help Neo Security understand security operations, buyer expectations and the operational weaknesses of existing products. It may also shorten customer access because the founders have relationships developed through earlier roles.

Founder reputation does not remove product-market risk. Neo Security is entering a category whose commercial boundaries remain unsettled, and experienced executives can still misjudge how quickly customers will allocate budgets or whether incumbent platforms can provide adequate functionality.

The management team must also shift from identifying a problem to building a scalable company. Product development, endpoint compatibility, international privacy obligations, customer support and channel strategy must progress together. A strong founder network can open doors, but retention and expansion will depend on whether security teams see measurable reduction in risk.

Can Neo establish a new category before larger cybersecurity platforms absorb agent control?

Neo Security is entering a market that every major cybersecurity provider is examining. CrowdStrike, Palo Alto Networks, SentinelOne, Microsoft and Cisco already control important parts of the endpoint, identity, cloud or security-operations stack. These companies can add agent discovery and policy enforcement to platforms that enterprises already license.

Incumbents have advantages in distribution, telemetry and procurement. A customer may prefer an additional module from an existing vendor rather than deploy another endpoint component, negotiate another contract and train security teams on a separate console.

Neo Security’s opportunity is that established platforms were built around older assumptions. Adding agent visibility to an endpoint product may not provide the same understanding of models, tools, delegated identities and multi-step actions as a system designed specifically for agentic software.

A focused startup can also move faster while the category is developing. Neo Security can adjust its inventory, policy language and control architecture as new agent frameworks, browser capabilities and Model Context Protocol services emerge. Larger companies may have more resources but must integrate changes across broader product portfolios and existing customer commitments.

The likely competitive battleground will be attribution. Enterprises need to know whether an action came from a person, an agent acting for that person, an autonomous service account or a sequence involving several tools. The provider that can build a reliable chain from identity to action may control a strategically important layer of enterprise AI governance.

Neo Security must avoid becoming only a visibility dashboard. Discovery products can generate initial interest, but long-term value will depend on whether the platform can enforce policies accurately without blocking legitimate work. False positives that repeatedly interrupt developers or business users could weaken adoption even when the underlying risk is real.

What commercial model could turn agent inventory and policy enforcement into recurring revenue?

Neo Security is likely to pursue an enterprise subscription model tied to endpoints, protected identities, agentic applications or a combination of usage measures. The company has not publicly disclosed pricing, so the eventual structure will be important in determining whether adoption expands with agent usage or becomes constrained by licensing complexity.

Endpoint-based pricing would be familiar to security buyers and relatively easy to budget. However, it may not reflect the number or risk of agents operating on each device. Usage-based pricing could capture greater value as agent activity expands, but unpredictable bills may discourage broad deployment.

A platform subscription could combine inventory, posture analysis and runtime control under tiered enterprise packages. This may support recurring revenue and product expansion, particularly if customers begin with visibility before activating enforcement across selected applications.

The strongest commercial path is likely to involve a phased deployment. Security teams may initially map agent exposure, identify excessive permissions and establish policies before allowing Neo Security to block actions in production. That progression creates an opportunity to convert assessment projects into broader subscriptions.

Channel partnerships could also matter. Managed security providers, technology integrators and endpoint vendors may help Neo Security reach companies without large internal security teams. However, partnerships can reduce retained revenue and may become difficult if potential partners develop competing agent-security products.

Neo Security will need to demonstrate value through operational outcomes. Useful measures could include the number of previously unknown agents discovered, excessive permissions corrected, sensitive actions blocked and investigation time reduced. Without measurable outcomes, agent security risks becoming another dashboard supported by a large budget and an impressive vocabulary.

Which technical and privacy risks could weaken Neo’s enterprise proposition?

The first challenge is accurate discovery. Agentic functionality can appear inside browsers, plug-ins, scripts, software-as-a-service applications and custom enterprise workflows. Neo Security must maintain visibility as new tools emerge and existing products change their behaviour through updates.

The second challenge is attribution. One action may involve a human instruction, an AI model, an agent framework, a service account and several external tools. Assigning responsibility to the wrong actor could produce ineffective policies or misleading audit records.

The third challenge is enforcement quality. Controls must stop genuinely risky actions without disrupting ordinary work. Policies that repeatedly hold legitimate exports, code changes or automated processes could encourage users to disable the product or seek workarounds.

Privacy is another consideration because endpoint visibility can expose employee activity, application usage and sensitive business data. Neo Security says its analysis can remain local to the device, but enterprises will still require clarity on what metadata leaves the endpoint, how policy decisions are recorded and who can access audit information.

Security of the platform itself will be critical. A tool with authority to observe identities, intercept actions and enforce endpoint policies becomes an attractive target. Neo Security must show strong software-supply-chain controls, update integrity and administrative protection before customers place it inside highly sensitive environments.

The company must also support diverse operating systems and enterprise configurations. Neo Security currently highlights Windows and macOS availability, while many corporate workloads and development environments also rely heavily on Linux, containers and cloud-native infrastructure. Coverage gaps could limit adoption unless the roadmap expands.

What evidence will show whether Neo’s $100 million launch is creating a durable security business?

The first measurable proof point will be named enterprise customers moving beyond testing into paid production deployments. Financial services, transportation and energy trials provide useful validation, but investors and competitors will look for organisations willing to publicly associate their security architecture with the platform.

The second test will be recurring revenue and retention. Neo Security must show that customers expand from discovery into policy enforcement and deploy the platform across larger endpoint populations rather than treating it as a limited assessment tool.

The third test will be technical coverage. Support for additional operating systems, applications, agent frameworks and identity platforms will indicate whether the company can control the wider enterprise environment rather than a narrow set of developer tools.

The fourth test will be policy accuracy. Neo Security needs evidence that it can prevent high-risk activity while maintaining application performance and producing an acceptable false-positive rate.

The fifth test will be competitive positioning. Integrations with established security platforms would suggest that Neo Security can become an independent control layer, while rapid imitation by incumbent vendors could compress its differentiation.

The $100 million funding round improves Neo Security’s ability to build ahead of the market, but it does not establish product leadership by itself. The company has identified a credible security gap created by software that can act rather than merely respond.

What has improved is the availability of capital, an experienced leadership team and a clearly defined product architecture. What remains unresolved is commercial adoption, pricing, customer retention and whether endpoint enforcement becomes the preferred method for controlling enterprise agents.

The next meaningful proof will not be another funding announcement. It will be evidence that large organisations are allowing Neo Security to govern real AI actions across production systems and are expanding those deployments after seeing measurable results.

Key takeaways on Neo’s $100 million AI agent security launch and enterprise strategy

• Neo Security emerged from stealth with $100 million in combined seed and Series A funding led by Andreessen Horowitz and Bessemer Venture Partners.
• The company is targeting software that gains autonomous capabilities through AI models, tools, browser extensions and enterprise application updates.
• Neo Security’s platform is designed to discover agentic software, identify permissions and data access, attribute actions and enforce runtime policies.
• The endpoint-based architecture may complement existing identity, data and threat-detection products rather than replacing the entire security stack.
• The founding team brings SentinelOne, EasySend, Wiz and Palo Alto Networks experience, improving commercial credibility and technical recruitment potential.
• Neo Security has not disclosed its valuation, revenue, customer count or the division of capital between its funding rounds.
• Testing across financial services, transportation and energy provides early validation, but named production customers remain the next important proof point.
• Established cybersecurity companies could integrate similar agent controls into broader platforms, increasing pressure on Neo Security to build differentiation quickly.
• Deployment success will depend on accurate discovery, reliable attribution, low false-positive rates and strong protection of endpoint and employee data.
• Durable value will require recurring revenue, customer expansion and evidence that Neo Security prevents harmful agent actions without obstructing legitimate enterprise work.