top of page

The $46 Million Arrakis Bet: Why AI Agents Are Becoming the Next Enterprise Battleground

Writer: Professor Matt Crump
Professor Matt Crump
Aug 3
8 min read
Artificial intelligence is moving from software that responds to human instructions toward autonomous agents capable of accessing systems, interpreting information, executing workflows, and making decisions with limited human intervention. That transition is creating a major enterprise opportunity, but it is also exposing a new category of cybersecurity and operational challenges. Two companies named Arrakis illustrate different sides of this transformation. Arrakis Security, founded by cybersecurity veterans from Torq and Palantir, is developing technology to monitor and govern the behavior of enterprise AI agents. Separately, London-based Arrakis is building an AI deployment platform designed to integrate autonomous agents into complex industrial workflows. Although the companies operate in different markets, their strategies point toward the same broader shift: AI agents are becoming participants in enterprise environments rather than merely tools used by employees. The Enterprise Is Entering the Agentic AI Era Traditional enterprise software generally operates according to predefined rules. Humans initiate actions, authenticate themselves, make decisions, and use applications to complete tasks. AI agents change this model by allowing software to interpret objectives and independently perform sequences of actions across multiple systems. An agent could, for example, retrieve information from a customer relationship management platform, analyze financial records, initiate a procurement workflow, update an internal database, and communicate the result to another application. This creates significant productivity potential, particularly in organizations burdened by legacy infrastructure, fragmented data, and repetitive manual processes. It also introduces a fundamental security problem. The question is no longer simply whether a user has permission to access a system. Enterprises increasingly need to determine whether an autonomous software identity is behaving appropriately after it receives that access. That distinction is at the center of Arrakis Security's strategy. Arrakis Security Targets the Behavioral Layer of AI Cybersecurity Israeli cybersecurity startup Arrakis Security has raised $8 million in Seed funding to develop infrastructure for monitoring and governing enterprise AI agents. The funding round was led by Hetz Ventures and included investors such as ElevenLabs CEO Mati Staniszewski, Torq CEO Ofer Smadari, Pentera founder and CEO Amitai Ratzzon, and senior Palantir executives. Arrakis Security was founded by Tal Baron, Omer Efrat, and Ron Shani, veterans of Torq and Palantir. The company employs approximately 20 people and intends to use its new capital to expand research, development, and product teams in Israel while improving detection, response, and enterprise deployment capabilities. Its central argument is that conventional cybersecurity architectures were built primarily around human users, endpoints, applications, and known identity structures. Autonomous agents introduce a fundamentally different behavioral model. An AI agent can potentially operate continuously, access multiple applications, react to changing information, and execute actions at a speed and scale that human employees cannot match. That creates what Arrakis describes as a visibility gap. An organization may know that an AI agent exists and may know which systems it can access, yet still lack a detailed understanding of what the agent is actually doing. Why AI Agent Behavior Requires a New Security Model Identity and access management remains important, but authorization alone cannot answer every question raised by autonomous AI. Consider an agent authorized to access a company's CRM system. Its access may be legitimate, but that does not automatically mean every action it performs is legitimate. Security teams increasingly need to understand: Which systems an AI agent is accessing What information it is retrieving Whether it is using expected permissions Which actions it is taking autonomously Whether it is transferring sensitive information Whether it is activating unexpected services Whether its behavior has deviated from established patterns Arrakis Security's platform is designed around this behavioral perspective. It seeks to discover AI agents throughout an organization, establish behavioral profiles, monitor activity in real time, and identify deviations that may represent security threats. When suspicious activity is detected, the system can potentially block actions, revoke permissions, suspend integrations, or activate predefined response workflows. This represents a movement from static AI security toward continuous behavioral governance. Non-Human Identities Are Becoming a Major Security Challenge The rise of AI agents also expands the importance of non-human identities. Modern enterprises already manage service accounts, automated processes, application credentials, bots, and machine-to-machine connections. AI agents can make these identities considerably more dynamic because they may be capable of deciding which actions to perform based on context. That creates a difficult governance problem. An enterprise security team must know not only who or what is connected to its infrastructure, but also why an automated system is acting, what it is attempting to accomplish, and whether its behavior remains within an approved operational boundary. Potential threats include unauthorized access to confidential information, privilege escalation, data exfiltration, misuse of credentials, and unexpected activation of enterprise services. The larger AI agents become within corporate workflows, the more consequential these risks become. Arrakis Security's establishment of Arrakis Labs, staffed by former Microsoft security researchers and technology-unit veterans, reflects the importance of research in this environment. Agent-specific vulnerabilities and attack techniques are likely to evolve alongside the technology itself. A Separate Arrakis Is Bringing AI Agents Into Industrial Operations The enterprise AI opportunity extends beyond cybersecurity. A separate London-based company also named Arrakis has raised approximately $38 million to accelerate deployment of AI agents across industrial operations. Its financing includes a $30 million Series A led by Blossom Capital, with participation from Accel and other investors. This Arrakis was founded in January 2026 by Rafael Quintanilla, Haroun Beltaifa, Romain Fouilland, and Mikhail Galkov. Its focus is markedly different from Arrakis Security. Rather than protecting AI agents, the company is building infrastructure to deploy them into complex industrial workflows. Its target industries include: Industry Potential AI Agent Applications Aerospace Workflow coordination, procurement, operational analysis Manufacturing Process automation, production intelligence Logistics Supply chain workflows and operational visibility Energy Data-driven operational processes Telecommunications Workflow management and system integration Construction Project and operational coordination The platform is designed to work with existing enterprise technology instead of requiring companies to replace their established software infrastructure. That is strategically important. Industrial organizations often operate environments containing legacy systems, specialized applications, disconnected databases, and processes that cannot easily be redesigned from scratch. AI Agents Could Become the Integration Layer for Legacy Industry One of the strongest arguments for enterprise AI agents is their ability to operate across fragmented systems. Instead of forcing employees to manually transfer information between applications, an agent can potentially coordinate tasks across those systems. This turns AI into an operational integration layer. The London-based Arrakis emphasizes model flexibility, allowing organizations to work with their own data while maintaining control over intellectual property. Its model-agnostic approach is also significant because the AI model landscape is changing rapidly. Companies that tie critical workflows too tightly to one model provider may face unnecessary technological dependence. A platform that can adapt as models improve could provide greater strategic flexibility. Early customer results described by the company indicate the potential economic impact. Some organizations have reportedly reduced procurement cycle times by as much as 90%, while others have gained real-time visibility into processes previously handled manually. Within six months of launch, the company had secured customers in Europe and the United States, including several NYSE-listed organizations in industrial, energy, and logistics markets. Security and Productivity Must Develop Together The two Arrakis strategies reveal an important principle about enterprise AI. AI deployment and AI security cannot be treated as separate problems. An industrial company may use autonomous agents to automate procurement, coordinate logistics, analyze operational data, or manage workflows. Those systems may create substantial efficiency gains, but their value depends on reliable governance. The more authority an agent receives, the greater the consequences of an error or compromise. This produces a fundamental enterprise equation: Greater autonomy = greater operational leverage + greater security responsibility. Organizations therefore need controls capable of monitoring agents throughout their lifecycle, from deployment and authentication to ongoing behavioral analysis and automated response. The Business Case for Agentic AI Is Becoming More Concrete The attraction of AI agents is not simply that they can generate text or summarize information. Their commercial potential comes from completing work. For enterprises, the most compelling use cases are likely to involve measurable outcomes such as: Reduced processing time Lower administrative overhead Faster procurement Improved operational visibility Automated coordination across applications Faster response to changing conditions More efficient use of specialized personnel The London-based Arrakis is taking an outcome-oriented approach, with a significant portion of its commercial model reportedly linked to customer value. That model could become increasingly important as enterprises become more skeptical of AI products that generate impressive demonstrations without delivering measurable financial or operational benefits. The New AI Stack Will Need Governance by Design The next generation of enterprise architecture is likely to include several interconnected layers. At the bottom sits the traditional infrastructure layer, including cloud systems, databases, enterprise applications, and operational technology. Above that sits the AI agent layer, where autonomous systems execute tasks and coordinate processes. A governance layer then becomes necessary to monitor identities, permissions, actions, data access, behavioral patterns, and policy compliance. This is where AI security platforms such as Arrakis Security could become strategically important. The emerging architecture can therefore be understood as: Infrastructure → AI agents → Autonomous actions → Behavioral monitoring → Policy enforcement → Automated response This is fundamentally different from conventional endpoint-centric security because the primary object of protection is no longer merely a device. It is an autonomous decision-making process. Challenges Could Determine How Quickly Enterprises Adopt AI Agents Despite the opportunity, autonomous enterprise AI faces substantial obstacles. Security is only one of them. Organizations must also address reliability, explainability, data governance, model errors, regulatory compliance, human oversight, integration complexity, and accountability. An AI agent that makes a mistake in a low-risk administrative workflow may create an inconvenience. The same type of failure in aerospace, energy, finance, manufacturing, or logistics could have considerably larger consequences. Enterprises will therefore need mechanisms to define what agents are allowed to do, what decisions require human approval, and what actions should trigger automatic intervention. The challenge is to achieve enough autonomy to generate meaningful productivity gains without creating uncontrolled operational risk. The Next Competitive Frontier Is Trustworthy Autonomy The AI industry is increasingly moving beyond the question of whether models can generate useful outputs. The more consequential question is whether autonomous systems can be trusted to operate inside real organizations. That requires two capabilities developing simultaneously: AI deployment platforms that can translate intelligence into measurable operational results, and security platforms capable of understanding and controlling what autonomous agents actually do. The two Arrakis companies occupy different positions within this emerging ecosystem, but both reflect the same technological transformation. For organizations considering agentic AI, the strategic objective should not be maximum autonomy at any cost. It should be controlled autonomy, where AI systems can act independently while remaining observable, governable, and accountable. What Arrakis Signals About the Future of Enterprise AI The rapid emergence of specialized platforms for both deploying and securing AI agents suggests that autonomous software is becoming an architectural category of its own. The next generation of enterprise technology will likely be shaped by systems that do not simply assist employees but actively execute portions of business operations. That shift creates enormous opportunities across industrial automation, cybersecurity, logistics, manufacturing, energy, aerospace, and enterprise software. It also makes behavioral visibility and governance indispensable. For analysts such as Dr. Shahid Masood and technology research organizations such as 1950.ai, the significance extends beyond individual startups. The larger story is the transition from generative AI toward autonomous digital infrastructure, where software increasingly acts as an independent participant in economic and operational systems. The companies that ultimately succeed will not necessarily be those that give AI the greatest freedom. They will be those that make autonomy useful, measurable, secure, and controllable. Key Takeaways Arrakis Security has raised $8 million in Seed funding to develop cybersecurity infrastructure for autonomous enterprise AI agents. Its platform focuses on agent discovery, behavioral profiling, real-time monitoring, non-human identities, and automated security response. A separate London-based Arrakis has raised approximately $38 million to deploy AI agents across industrial workflows. The industrial platform targets aerospace, manufacturing, logistics, energy, telecommunications, and construction. Early reported customers have achieved significant operational improvements, including procurement cycle reductions of up to 90%. Both companies illustrate the growing importance of autonomous AI in enterprise environments. As AI agents receive greater access and authority, security, governance, observability, and accountability will become essential components of enterprise AI architecture. Further Reading / External References Former Palantir and Torq veterans raise $8 million Seed to secure the rise of AI agents Arrakis Raises $38M to Scale AI Deployment for Industrial Operations

Artificial intelligence is moving from software that responds to human instructions toward autonomous agents capable of accessing systems, interpreting information, executing workflows, and making decisions with limited human intervention. That transition is creating a major enterprise opportunity, but it is also exposing a new category of cybersecurity and operational challenges.


Two companies named Arrakis illustrate different sides of this transformation. Arrakis

Security, founded by cybersecurity veterans from Torq and Palantir, is developing technology to monitor and govern the behavior of enterprise AI agents. Separately, London-based Arrakis is building an AI deployment platform designed to integrate autonomous agents into complex industrial workflows.

Although the companies operate in different markets, their strategies point toward the same broader shift: AI agents are becoming participants in enterprise environments rather than merely tools used by employees.


The Enterprise Is Entering the Agentic AI Era

Traditional enterprise software generally operates according to predefined rules. Humans initiate actions, authenticate themselves, make decisions, and use applications to complete tasks. AI agents change this model by allowing software to interpret objectives and independently perform sequences of actions across multiple systems.

An agent could, for example, retrieve information from a customer relationship management platform, analyze financial records, initiate a procurement workflow, update an internal database, and communicate the result to another application.

This creates significant productivity potential, particularly in organizations burdened by legacy infrastructure, fragmented data, and repetitive manual processes. It also introduces a fundamental security problem.


The question is no longer simply whether a user has permission to access a system. Enterprises increasingly need to determine whether an autonomous software identity is behaving appropriately after it receives that access.

That distinction is at the center of Arrakis Security's strategy.


Arrakis Security Targets the Behavioral Layer of AI Cybersecurity

Israeli cybersecurity startup Arrakis Security has raised $8 million in Seed funding to develop infrastructure for monitoring and governing enterprise AI agents.

The funding round was led by Hetz Ventures and included investors such as ElevenLabs CEO Mati Staniszewski, Torq CEO Ofer Smadari, Pentera founder and CEO Amitai Ratzzon, and senior Palantir executives.

Arrakis Security was founded by Tal Baron, Omer Efrat, and Ron Shani, veterans of Torq and Palantir. The company employs approximately 20 people and intends to use its new capital to expand research, development, and product teams in Israel while improving detection, response, and enterprise deployment capabilities.


Its central argument is that conventional cybersecurity architectures were built primarily around human users, endpoints, applications, and known identity structures. Autonomous agents introduce a fundamentally different behavioral model.

An AI agent can potentially operate continuously, access multiple applications, react to changing information, and execute actions at a speed and scale that human employees cannot match.

That creates what Arrakis describes as a visibility gap. An organization may know that an AI agent exists and may know which systems it can access, yet still lack a detailed understanding of what the agent is actually doing.


Why AI Agent Behavior Requires a New Security Model

Identity and access management remains important, but authorization alone cannot answer every question raised by autonomous AI.

Consider an agent authorized to access a company's CRM system. Its access may be legitimate, but that does not automatically mean every action it performs is legitimate.

Security teams increasingly need to understand:

  • Which systems an AI agent is accessing

  • What information it is retrieving

  • Whether it is using expected permissions

  • Which actions it is taking autonomously

  • Whether it is transferring sensitive information

  • Whether it is activating unexpected services

  • Whether its behavior has deviated from established patterns

Arrakis Security's platform is designed around this behavioral perspective. It seeks to discover AI agents throughout an organization, establish behavioral profiles, monitor activity in real time, and identify deviations that may represent security threats.

When suspicious activity is detected, the system can potentially block actions, revoke permissions, suspend integrations, or activate predefined response workflows.

This represents a movement from static AI security toward continuous behavioral

governance.


Non-Human Identities Are Becoming a Major Security

Challenge

The rise of AI agents also expands the importance of non-human identities.

Modern enterprises already manage service accounts, automated processes, application credentials, bots, and machine-to-machine connections. AI agents can make these identities considerably more dynamic because they may be capable of deciding which actions to perform based on context.

That creates a difficult governance problem.


An enterprise security team must know not only who or what is connected to its infrastructure, but also why an automated system is acting, what it is attempting to accomplish, and whether its behavior remains within an approved operational boundary.

Potential threats include unauthorized access to confidential information, privilege escalation, data exfiltration, misuse of credentials, and unexpected activation of enterprise services.

The larger AI agents become within corporate workflows, the more consequential these risks become.

Arrakis Security's establishment of Arrakis Labs, staffed by former Microsoft security researchers and technology-unit veterans, reflects the importance of research in this environment. Agent-specific vulnerabilities and attack techniques are likely to evolve

alongside the technology itself.


A Separate Arrakis Is Bringing AI Agents Into Industrial Operations

The enterprise AI opportunity extends beyond cybersecurity.

A separate London-based company also named Arrakis has raised approximately $38 million to accelerate deployment of AI agents across industrial operations. Its financing includes a $30 million Series A led by Blossom Capital, with participation from Accel and other investors.

This Arrakis was founded in January 2026 by Rafael Quintanilla, Haroun Beltaifa, Romain Fouilland, and Mikhail Galkov.

Its focus is markedly different from Arrakis Security. Rather than protecting AI agents, the company is building infrastructure to deploy them into complex industrial workflows.

Its target industries include:

Industry

Potential AI Agent Applications

Aerospace

Workflow coordination, procurement, operational analysis

Manufacturing

Process automation, production intelligence

Logistics

Supply chain workflows and operational visibility

Energy

Data-driven operational processes

Telecommunications

Workflow management and system integration

Construction

Project and operational coordination

The platform is designed to work with existing enterprise technology instead of requiring companies to replace their established software infrastructure.

That is strategically important. Industrial organizations often operate environments containing legacy systems, specialized applications, disconnected databases, and processes that cannot easily be redesigned from scratch.


AI Agents Could Become the Integration Layer for Legacy Industry

One of the strongest arguments for enterprise AI agents is their ability to operate across fragmented systems.

Instead of forcing employees to manually transfer information between applications, an agent can potentially coordinate tasks across those systems. This turns AI into an operational integration layer.

The London-based Arrakis emphasizes model flexibility, allowing organizations to work with their own data while maintaining control over intellectual property. Its model-agnostic approach is also significant because the AI model landscape is changing rapidly.

Companies that tie critical workflows too tightly to one model provider may face unnecessary technological dependence. A platform that can adapt as models improve could provide greater strategic flexibility.

Early customer results described by the company indicate the potential economic impact. Some organizations have reportedly reduced procurement cycle times by as much as 90%, while others have gained real-time visibility into processes previously handled manually.

Within six months of launch, the company had secured customers in Europe and the United States, including several NYSE-listed organizations in industrial, energy, and logistics markets.


Security and Productivity Must Develop Together

The two Arrakis strategies reveal an important principle about enterprise AI.

AI deployment and AI security cannot be treated as separate problems.

An industrial company may use autonomous agents to automate procurement, coordinate logistics, analyze operational data, or manage workflows. Those systems may create substantial efficiency gains, but their value depends on reliable governance.

The more authority an agent receives, the greater the consequences of an error or compromise.

This produces a fundamental enterprise equation:

Greater autonomy = greater operational leverage + greater security responsibility.

Organizations therefore need controls capable of monitoring agents throughout their lifecycle, from deployment and authentication to ongoing behavioral analysis and automated response.


The Business Case for Agentic AI Is Becoming More Concrete

The attraction of AI agents is not simply that they can generate text or summarize information. Their commercial potential comes from completing work.

For enterprises, the most compelling use cases are likely to involve measurable outcomes such as:

  1. Reduced processing time

  2. Lower administrative overhead

  3. Faster procurement

  4. Improved operational visibility

  5. Automated coordination across applications

  6. Faster response to changing conditions

  7. More efficient use of specialized personnel

The London-based Arrakis is taking an outcome-oriented approach, with a significant portion of its commercial model reportedly linked to customer value.

That model could become increasingly important as enterprises become more skeptical of AI products that generate impressive demonstrations without delivering measurable financial or operational benefits.


The New AI Stack Will Need Governance by Design

The next generation of enterprise architecture is likely to include several interconnected layers.

At the bottom sits the traditional infrastructure layer, including cloud systems, databases, enterprise applications, and operational technology.

Above that sits the AI agent layer, where autonomous systems execute tasks and coordinate processes.

A governance layer then becomes necessary to monitor identities, permissions, actions, data access, behavioral patterns, and policy compliance.

This is where AI security platforms such as Arrakis Security could become strategically important.

The emerging architecture can therefore be understood as:

Infrastructure → AI agents → Autonomous actions → Behavioral monitoring → Policy enforcement → Automated response

This is fundamentally different from conventional endpoint-centric security because the primary object of protection is no longer merely a device. It is an autonomous decision-making process.


Challenges Could Determine How Quickly Enterprises Adopt AI Agents

Despite the opportunity, autonomous enterprise AI faces substantial obstacles.

Security is only one of them. Organizations must also address reliability, explainability, data governance, model errors, regulatory compliance, human oversight, integration complexity, and accountability.

An AI agent that makes a mistake in a low-risk administrative workflow may create an inconvenience. The same type of failure in aerospace, energy, finance, manufacturing, or logistics could have considerably larger consequences.

Enterprises will therefore need mechanisms to define what agents are allowed to do, what decisions require human approval, and what actions should trigger automatic intervention.

The challenge is to achieve enough autonomy to generate meaningful productivity gains without creating uncontrolled operational risk.


The Next Competitive Frontier Is Trustworthy Autonomy

The AI industry is increasingly moving beyond the question of whether models can generate useful outputs.

The more consequential question is whether autonomous systems can be trusted to operate inside real organizations.


That requires two capabilities developing simultaneously: AI deployment platforms that can translate intelligence into measurable operational results, and security platforms capable of understanding and controlling what autonomous agents actually do.

The two Arrakis companies occupy different positions within this emerging ecosystem, but both reflect the same technological transformation.

For organizations considering agentic AI, the strategic objective should not be maximum autonomy at any cost. It should be controlled autonomy, where AI systems can act independently while remaining observable, governable, and accountable.


What Arrakis Signals About the Future of Enterprise AI

The rapid emergence of specialized platforms for both deploying and securing AI agents suggests that autonomous software is becoming an architectural category of its own.

The next generation of enterprise technology will likely be shaped by systems that do not simply assist employees but actively execute portions of business operations.

That shift creates enormous opportunities across industrial automation, cybersecurity, logistics, manufacturing, energy, aerospace, and enterprise software. It also makes behavioral visibility and governance indispensable.


For analysts such as Dr. Shahid Masood and technology research organizations such as

1950.ai, the significance extends beyond individual startups. The larger story is the transition from generative AI toward autonomous digital infrastructure, where software increasingly acts as an independent participant in economic and operational systems.

The companies that ultimately succeed will not necessarily be those that give AI the greatest freedom. They will be those that make autonomy useful, measurable, secure, and controllable.


Key Takeaways

  • Arrakis Security has raised $8 million in Seed funding to develop cybersecurity infrastructure for autonomous enterprise AI agents.

  • Its platform focuses on agent discovery, behavioral profiling, real-time monitoring, non-human identities, and automated security response.

  • A separate London-based Arrakis has raised approximately $38 million to deploy AI agents across industrial workflows.

  • The industrial platform targets aerospace, manufacturing, logistics, energy, telecommunications, and construction.

  • Early reported customers have achieved significant operational improvements, including procurement cycle reductions of up to 90%.

  • Both companies illustrate the growing importance of autonomous AI in enterprise environments.

  • As AI agents receive greater access and authority, security, governance, observability, and accountability will become essential components of enterprise AI architecture.


Comments

bottom of page