Author: Techtime

Mobileye Develops a “Social Intelligence” Layer for Autonomous Driving Systems

Posted on by Techtime

By Yohai Schwiger

During its earnings call yesterday, following the release of its financial results, Mobileye CEO Amnon Shashua revealed that the company is developing a new intelligence layer for its autonomous driving systems. The goal, he explained, is to add a level of understanding above the system’s immediate decision-making layer — the one responsible for analyzing the environment, identifying open lanes, and computing trajectories in real time. Shashua described the shift as a fundamental change in how the system understands the driving environment, stressing that “autonomous driving is not just a geometric perception problem, but a decision-making problem in a social environment.”

Shashua framed the road as a space populated by many independent agents, each making decisions that influence the behavior of others. Every driver, pedestrian, cyclist, or traffic actor is a participant in a single, interconnected system. A single action by an autonomous vehicle alters the behavior of the surrounding environment, and those reactions, in turn, feed back into the vehicle’s own decision-making. In such conditions, Shashua argued, understanding the road requires a broader perspective than simply identifying an open path or calculating an optimal trajectory.

Billions of Simulation Hours Overnight

Against this backdrop, Shashua introduced the concept of Artificial Community Intelligence, or ACI. According to him, the idea originates in academic research, but has never before been implemented at commercial scale in autonomous driving systems. “This is the first time an academic concept has been fully productized in an autonomous vehicle decision-making system,” he said. ACI is built on multi-agent reinforcement learning, aimed at understanding the dynamics of a community. The core idea is to train the system to understand how one decision affects the behavior of others, and how that chain of reactions ultimately feeds back into the vehicle itself.

This is a dimension familiar to every human driver, even if they are rarely conscious of it. A driver approaching a busy roundabout or a school pickup zone does not merely calculate available space or distance to the car ahead. Instead, they interpret the entire situation: children standing on the sidewalk, parents double-parked, impatient drivers nearby, and a general sense that the space is operating under different social rules. The decision to slow down, yield, or wait an extra moment is not derived from geometry alone, but from a deeper understanding of context and expectations.

To train such capabilities, Mobileye makes extensive use of simulation. Shashua explained that while visual perception can be trained using real-world data, planning and decision-making in a multi-agent environment require far larger volumes of data. “The sample complexity for planning is significantly higher,” he said, “because the actions you take influence the behavior of others.” The solution, according to Shashua, is large-scale simulation that allows the company to run massive numbers of scenarios. “We can reach a billion training hours overnight,” he said, highlighting the advantage of combining simulation with Mobileye’s global REM mapping infrastructure, which provides a realistic foundation for training.

Thinking Fast, Thinking Slow

The insights learned through simulation are divided by Mobileye into two cognitive layers: Fast Think and Slow Think. The Fast Think layer is responsible for real-time actions and manages the vehicle’s immediate safety. It operates at high frequencies — dozens of times per second — handling steering, braking, and lane keeping. This is a reflexive layer that runs directly on the vehicle’s hardware and cannot tolerate delays or uncertainty.

By contrast, the Slow Think layer focuses on understanding the situation in which the vehicle is operating. It does not ask only what is permitted or prohibited, but what is appropriate in a given context. Shashua described it as a layer designed to interpret the meaning of complex, non-routine scenes, rather than merely avoiding immediate hazards. For example, when a police officer blocks a lane and signals a vehicle to wait or change course, a safety system will know not to hit the officer. A contextual understanding layer, however, allows the vehicle to recognize that this is not a typical obstacle, but a human directive that requires a change in behavior.

The distinction between Fast Think and Slow Think closely echoes the fast-and-slow thinking model popularized by Nobel Prize-winning economist and psychologist Daniel Kahneman, who coined the terms to describe the difference between reflexive, automatic thinking and slower, interpretive reasoning. While Shashua did not mention Kahneman by name, the conceptual parallel is clear.

Architecturally, this translates into a clear separation between real-time systems and interpretive systems. Fast Think operates within tight control loops, while Slow Think runs at a lower cadence, providing context and high-level guidance. Its output is not a steering or braking command, but a shift in driving policy. It may cause the system to adopt more conservative behavior, avoid overtaking, or prefer yielding the right of way. The planner continues to compute trajectories in real time, but does so under a new set of priorities.

Shashua noted that because the Slow Think layer is not safety-critical and does not operate in real time, it is not bound by the same hardware constraints as the execution layer. In principle, such processing could even be performed using heavier compute resources, including the cloud, to analyze complex situations more deeply. He emphasized, however, that this does not involve moving driving decisions to the cloud, but rather expanding the system’s ability to understand situations that do not require immediate response.

Will This Extend to Mentee Robotics’ Humanoids?

The connection between ACI and the Slow Think layer is most evident during training. Simulation does not operate while the vehicle is driving, but serves as a laboratory in which the system learns social dynamics. In real time, the vehicle does not recompute millions of scenarios, but instead recognizes situations it has encountered before and applies intuition acquired through prior training. In this sense, simulation is where understanding is built, and Slow Think is where it is expressed.

The concepts Shashua outlined are not necessarily limited to the automotive domain. They may also be relevant to Mobileye’s recent expansion into humanoid robotics through its acquisition of Mentee Robotics. Like autonomous vehicles, humanoid robots operate in multi-agent environments, where humans and other machines react to one another in real time. Beyond motor control and balance, such robots must understand context, intent, and social norms. Mobileye has noted that Mentee’s robot is designed to learn by observing behavior around it, rather than relying solely on preprogrammed actions — an approach that aligns naturally with ideas such as ACI and the separation between immediate reaction and higher-level understanding.

Why Emphasize This in an Earnings Call?

This raises an obvious question: why did Shashua choose to devote so much time during an investor call to a deep technical topic focused on internal AI architecture, rather than on revenues, forecasts, or new contracts? The choice appears deliberate. On one level, it draws a clear line between Mobileye and approaches that frame autonomous driving as a problem solvable through technological shortcuts or brute-force increases in compute. Shashua sought to emphasize that the core challenge lies in understanding human interaction, not merely in object detection or trajectory planning. At the same time, the discussion serves as expectation management, explaining why truly advanced autonomous systems take time to develop and do not immediately translate into revenue. On a deeper level, it signals a broader strategy: Mobileye is not just building products or chips, but an intelligence layer for physical AI — one that could extend beyond the current generation of autonomous vehicles into adjacent domains such as robotics.

In that sense, the most technical segment of the earnings call was also the clearest business message. Not a promise for the next quarter, but an explanation of why the longer, more complex path is, in Mobileye’s view, the right one.

Claroty Raises $150 Million as Company Valuation Jumps 80%

Posted on by Techtime

[Photo: Claroty CEO Yaniv Vardi. Credit: Keren Mazor]

Israeli cybersecurity company Claroty, which develops technologies to protect critical infrastructure and cyber-physical systems (CPS), announced that it has raised $150 million in a Series F funding round. The round was led by Golub Growth, the growth investment arm of U.S.-based private credit firm Golub Capital, with participation from existing investors.

The funding will be used to accelerate the company’s global growth and expand its presence across critical infrastructure markets, amid rising cyber threats targeting industrial systems, energy, water, transportation, and healthcare. The company said the new round reflects an 80% increase in its valuation compared to its previous funding round in March 2024, when Claroty raised approximately $100 million at an estimated valuation of $2–2.5 billion. Based on this increase, the current round implies an estimated valuation of $3.6–4.5 billion.

Claroty operates in the field of cyber-physical systems protection—environments that combine software with physical processes and infrastructure, including industrial facilities, energy networks, water systems, transportation, and healthcare. The company’s platform is designed to provide organizations with deep, accurate visibility into operational assets, alongside exposure management, threat detection, network protection, and secure remote access. It is built specifically for OT environments and critical infrastructure, where traditional IT security solutions often fall short.

Over the past year, Claroty expanded its technology offering with the launch of CPS Library, an AI-powered catalog of physical assets aimed at improving asset identification, classification, and monitoring across cyber-physical environments. The library uses artificial intelligence to analyze partial and unstructured operational data, correlate assets with known vulnerabilities and cyber risks, and deliver a continuously updated, accurate view of critical assets—forming a foundation for risk management, regulatory compliance, and operational resilience.

During the past year, Claroty added 24 new Fortune 500 customers, and its platform received recognition from leading U.S. research firms, including Gartner and Forrester, which ranked the company’s solutions among the market leaders in CPS protection.

Yaniv Vardi, CEO of Claroty, said the new funding round marks a new chapter in the company’s growth. “This round marks the beginning of an exciting new phase in Claroty’s growth journey, as we continue building a purpose-built platform for protecting cyber-physical systems,” Vardi said. “Organizations operating in industrial environments need the right combination of technology, people, and processes to serve as a stable foundation for reducing risk, meeting regulatory requirements, and maintaining operational continuity.”

Claroty is headquartered in New York and operates across Europe, the Asia-Pacific region, and Latin America. The company serves hundreds of organizations across thousands of sites worldwide, primarily in critical infrastructure and industrial sectors.

Zeroport Develops Non-IP Remote Access System for Critical Infrastructure

Posted on by Techtime

[Image: Zeroport’s hardware-based access solution. Credit: Zeroport]

Israeli industrial cybersecurity company Zeroport has emerged from stealth today, announcing a $10 million Seed funding round. The round was led by lool ventures, with participation from Clarim Ventures, Elron Ventures’ micro-fund, CyberFuture, and Fusion Fund. The company plans to use the funding to support global expansion, grow its workforce, and continue developing its core platform.

Zeroport operates in the secure remote access space, addressing one of the most sensitive challenges in modern cybersecurity: enabling remote connectivity to critical and highly sensitive systems without exposing them to the internet or to IP-based communication. Since its founding, the company has already deployed its technology among large organizations in Israel and internationally, including operators of critical infrastructure, energy companies, financial institutions, as well as government and defense entities.

In recent years, remote access to operational and critical systems has become unavoidable. Organizations now manage geographically distributed facilities, rely heavily on external vendors and contractors, require real-time monitoring and maintenance, and must operate under emergency conditions where physical access to sites is impractical or impossible. Despite this reality, most organizations still depend on software-based solutions, primarily VPNs and various Zero Trust platforms—a security model that removes implicit trust and requires continuous authentication and authorization for every access attempt.

These solutions are built on IP communication and, by design, introduce an attack surface that can be exploited through malware, credential theft, or network intrusion.

This challenge is far from theoretical. Even organizations with advanced security postures continue to suffer breaches originating from remote access infrastructures. As a result, many are forced to choose between completely isolating sensitive systems from connectivity or relying on complex, costly, and inherently vulnerable remote access solutions. This dilemma is particularly acute in critical infrastructure and operational technology environments, where any network connection can become a single point of failure.

Zeroport’s technology takes a fundamentally different approach. Instead of relying on standard networking protocols or IP-based communication, user input and display output are converted into analog signals and transmitted via a dedicated hardware component. This architecture eliminates any direct IP communication between the remote user and the organization’s internal network.

According to the company, the complete absence of IP connectivity inherently prevents intrusion, interception, or malicious code injection, enabling secure remote access even to systems that were previously considered too sensitive to connect externally. The patented hardware component, installed at the organization’s perimeter, functions as a physical isolation layer between external users and critical internal systems, without relying on software-based security mechanisms.

Zeroport founders Yosef “Sefi” Gertz and Lavi Friedman. Credit: Zeroport

Zeroport was founded in mid-2024 in Herzliya by CEO and co-founder Yosef “Sefi” Gertz, who brings over 15 years of experience in leading technology companies, and co-founder and CTO Lavi Friedman, a graduate of Israel Defense Forces Unit 81. The idea for the company emerged during reserve military service in the “Iron Swords” war, where the founders identified deep structural gaps in existing secure remote access solutions.

Today, Zeroport employs approximately 25 people, with offices in Israel and additional locations in New York and Singapore. Following the funding round, the company plans to grow its workforce to around 40 employees within a year and expand into markets across North America, Europe, and the Asia-Pacific region.

Zeroport currently offers three core products, each designed to address different secure remote access needs while maintaining strict isolation from IP communication. Fantom is the company’s flagship, enterprise-scale solution, supporting large networks with multiple users—including employees, contractors, and third-party vendors—through dedicated hardware devices that create one-way isolation with no digital attack surface. Polter is designed for highly sensitive, one-to-one access scenarios, enabling secure, isolated connectivity to a single system or workstation. Roamer is a zero-client endpoint with no operating system, providing a hardened remote workspace that connects to Fantom or Polter without leaving data, memory, or attack vectors on the user’s device.

Zeroport is primarily targeting organizations that operate critical systems, including operational technology environments, industrial control systems, energy infrastructure, sensitive financial facilities, and government and defense bodies. In these environments, the technology enables secure remote maintenance, monitoring, and operation of sites and systems without requiring on-site personnel or exposing internal networks to the internet.

Beyond OT environments, the platform can also serve IT organizations that demand extreme levels of security and seek to replace layered, legacy remote access architectures accumulated over years.

According to the company, its solution is already in use among large organizations in Israel and abroad across critical infrastructure, energy, finance, and public sector segments. Early adoption has been driven by organizations with near-zero tolerance for error, where any remote access breach could result in severe operational, regulatory, or security consequences.

Trophy Selected as Baseline Protection for NATO’s Next-Generation Tanks

Posted on by Techtime

[Image: The Leopard 2A8 main battle tank. Photo credit: EuroTrophy GmbH and KNDS Deutschland]

Rafael’s Trophy active protection system has been selected as part of the standard baseline configuration of the Leopard 2A8, the next-generation variant of Germany’s Leopard 2 main battle tank family. The decision was made under a broad, multinational contract signed between EuroTrophy and tank manufacturer KNDS Deutschland, covering tank deliveries to four NATO members: Lithuania, the Netherlands, the Czech Republic and Croatia.

The contract, estimated at approximately €330 million, includes the supply of Trophy systems, spare parts, training, and comprehensive lifecycle logistical support. Beyond its financial scope, the agreement reflects a broader shift in how NATO defines baseline requirements for next-generation armored platforms.

Not an Upgrade — Integration from the Start

One of the most significant aspects of the deal is the way Trophy is being incorporated: not as a retrofit or optional add-on, but as an integral part of the tank’s standard baseline configuration.

This means the active protection system is embedded at the production stage, fully integrated into the tank’s electrical, sensor and command-and-control architecture, as well as its overall operational concept. The fact that KNDS Deutschland, the tank’s manufacturer, is a direct party to the agreement further reinforces the assessment that this is a production-line integration rather than a post-delivery modification. From NATO’s perspective, active protection is increasingly being defined as a baseline requirement rather than a future enhancement.

Leopard 2A8: A Veteran Platform, Reimagined

The Leopard 2 family entered service in the late 1970s, but the Leopard 2A8 represents a significant leap forward. It is the most advanced variant of the platform to date and the first in many years to be produced as a newly built tank, rather than a conversion of older hulls.

The 2A8 is designed for the contemporary battlefield — one characterized by advanced anti-tank missiles, drones, dense sensor networks and network-centric warfare. It combines an upgraded 120mm main gun, enhanced fire-control systems, multi-layered survivability and the ability to operate as part of a joint, multinational force.

Integrating Trophy as part of the baseline configuration complements this concept, providing real-time interception of anti-tank threats while significantly improving crew situational awareness.

How Many Tanks — and How Many Systems?

Based on publicly reported procurement plans, the four participating countries have each committed to purchasing dozens of Leopard 2A8 tanks, with a combined baseline total of approximately 170–180 tanks, and the potential for further expansion should options be exercised.

Since Trophy is defined as part of the tank’s baseline configuration, this implies the integration of several dozen active protection systems — potentially approaching 200 units if procurement plans are fully realized.

EuroTrophy: Bridging Israeli Technology and European Industry

The European implementation of the deal is being carried out through EuroTrophy, a German joint venture between Rafael, KNDS Deutschland and General Dynamics, headquartered in Frankfurt. The company focuses on the marketing, production and integration of the Trophy system for NATO and European Union customers.

This structure enables the incorporation of combat-proven Israeli technology into Europe’s defense-industrial base, while meeting local regulatory, industrial and logistical requirements — positioning Trophy as a core element of Europe’s armored warfare architecture for years to come.

Trophy: Operational Active Protection

Trophy, developed and manufactured by Rafael, is an operational active protection system designed to intercept anti-tank threats using a hard-kill mechanism. The system incorporates detection, tracking and alert capabilities, providing armored crews with information on the direction and source of incoming threats as part of the engagement cycle.

The system is deployed on Merkava tanks, M1 Abrams, Leopard 2 platforms and additional armored vehicles. Its selection as a baseline configuration for the Leopard 2A8 reflects the integration of active protection as an inherent component of modern Western armored force doctrine.

More Than a Contract — A Strategic Statement

Beyond its commercial dimensions, the agreement reflects lessons learned from recent battlefields and a growing recognition within NATO that tank survivability can no longer rely on passive armor alone. The shift toward integrated active protection marks a deeper transformation in armored warfare doctrine — one that positions the Leopard 2A8, together with Trophy, as a central pillar of Europe’s armored forces in the coming decade.

Hailo and ASUS Unveil USB-Based AI Accelerator for Personal Computers

Posted on by Techtime

[Pictured above: The ASUS UGen300 AI drive powered by a Hailo processor]

Israeli AI chipmaker Hailo and PC manufacturer ASUS have unveiled a new AI accelerator that connects to personal computers via USB, designed to support both traditional AI workloads and generative AI applications at the edge. The device, dubbed the ASUS UGen300, is the world’s first AI accelerator to arrive in a USB form factor, allowing users to connect advanced AI processing to a computer much like a standard external storage drive.

The USB-based accelerator enables edge AI capabilities on virtually any computing device, effectively turning it into an AI-ready workstation. By offloading AI workloads from the host system, the device allows developers and enterprises to run inference and generative AI tasks locally, without relying on cloud resources.

The UGen300 is powered by Hailo’s Hailo-10H processor, delivering up to 40 tera-operations per second (TOPS) for neural network acceleration. It integrates 8GB of LPDDR4 memory, enabling on-device generative AI performance using the drive’s dedicated memory, thereby avoiding additional load on the host computer. The plug-and-play USB-C connection supports Windows, Linux, and Android operating systems.

According to ASUS, the device supports a broad range of AI models, including large language models (LLMs), vision-language models (VLMs), Whisper, and a variety of computer vision networks. Users gain access to more than 100 pre-trained models through an online model zoo, spanning language, multimodal, and vision-based AI applications.

Average power consumption for the new AI drive is approximately 2.5 watts, making it suitable for continuous use on laptops and edge devices. The accelerator supports both x86 and ARM host architectures and can also be used to speed up traditional compute workloads, not limited solely to AI tasks.

Hailo announced the Hailo-10H processor around six months ago, in July 2025. It represents the company’s second generation of edge AI accelerators and is positioned as the first of its kind to enable local execution of generative AI models directly on the device, without requiring a cloud connection. The new chip builds on the vision AI capabilities introduced with Hailo’s first-generation Hailo-8 processor.

The Hailo-10H is fully integrated into Hailo’s existing development environment, which includes an active developer community of approximately 10,000 monthly users. The combination of a USB form factor, low power consumption, and broad model support positions the ASUS UGen300 as a potential enabler for bringing generative AI capabilities closer to users, at the edge of the network.

NewPhotonics Unveils Optical Communication Chip Designed for Near-Packaged Optics

Posted on by Techtime

By Yohai Schwiger

Israeli photonics company NewPhotonics, based in Petah Tikva, today unveiled its latest optical communication transmitter chip, the NPC50503 NPO, delivering data rates of up to 1.6 terabits per second (1.6 Tbps). The device is packaged in a low-power flip-chip BGA form factor and supports per-lane speeds of 224 Gbps, compliant with the PAM4 IEEE 802.3dj standard. It is designed for deployment in Near-Packaged Optics (NPO) architectures, which are increasingly required in large-scale AI data centers.

The new chip integrates an on-chip laser, an optical signal-processing unit, and an ultra-high-speed electrical interface. According to the company, this design reduces signal loss, lowers power consumption, and enables optical connectivity to be placed closer to processors and accelerators inside the server itself. The announcement also marks a strategic shift for NewPhotonics. Until now, the company’s products were primarily focused on enhancing and upgrading existing optical communication modules.

NewPhotonics previously supplied Transmitter-on-Chip (TOC) solutions, which replace the internal transmission components of optical modules and enable data rates of up to 1.6 Tbps while improving energy efficiency. This approach allowed module manufacturers to significantly boost performance without redesigning the underlying network architecture.

Moving Optical Links Closer to the Compute Core

The newly announced chip represents a move toward a more comprehensive solution positioned at the heart of AI-intensive systems. It is an NPO component designed to be installed inside the package, close to the CPU, GPU, or custom ASIC. This marks a meaningful architectural change. Instead of transmitting electrical signals from the processor to a photonic module via copper traces, which limit bandwidth and energy efficiency, optical connectivity is brought much closer to the point of computation.

In traditional systems, optical components reside in pluggable transceiver modules mounted at the front of servers or switches. These modules convert electrical signals to optical signals and back, but the electrical signal must first travel tens of centimeters over copper traces on the circuit board. At speeds exceeding 100 Gbps per lane, every additional centimeter degrades signal integrity, increases power consumption, and constrains overall bandwidth.

Near-Packaged Optics shortens this distance dramatically, sometimes to just a few centimeters, by integrating optical components on the board adjacent to the compute device itself. Reducing the electrical path enables higher data rates and better energy efficiency, while optical fibers become an integral part of the system architecture.

Refining the Optical Fabric for AI Workloads

Modern data centers, particularly those designed to run large AI models, require massive data movement between compute units while maintaining tight power budgets. Across the industry, there is a clear shift away from copper-based interconnects and discrete optical modules toward optics that are integrated directly into the system.

The next stage of this evolution is Co-Packaged Optics (CPO), where optical components and logic are combined within the same hardware package at the ASIC level. While CPO represents the long-term goal, it remains challenging due to manufacturing complexity, cost, and operational concerns, as a failure in an optical component could disable an entire package.

NewPhotonics’ NPO solution is positioned as a critical intermediate step toward that future. By integrating optical communication inside the system package rather than relying on external pluggable modules, the approach reduces dependence on copper, improves power efficiency, and still preserves design flexibility and operational practicality.

“We Want to Become Israel’s Largest Data Company”

Posted on by Techtime

[Pictured, right to left: Erez Moskovitz and Yoav Ziv. Photo credit: Eyal Taug]

Two Israeli technology companies, Tasq AI and BLEND, have announced a merger into a single entity operating under the Tasq AI brand, in a move designed to tackle one of the central challenges of the enterprise AI revolution: the growing trust gap between the impressive capabilities of AI models and organizations’ willingness to deploy them in mission-critical operations. The merger brings together a data and AI quality management platform with a global network of human experts, aiming to create a trust layer that allows enterprises to train and operate models and AI agents with production-grade reliability.

The logic behind the combination is relatively straightforward. Over the years, Tasq AI has built infrastructure for managing AI data across the entire lifecycle, from model training to validating outputs in production. BLEND, which operated in the translation and localization space, brings deep experience with a model in which automation performs most of the work, while humans step into the loop to refine, verify and preserve meaning. Together, the companies are seeking to extend this approach far beyond translation into a broad range of enterprise AI use cases.

“Today, the real problem is no longer the AI models themselves, but the data and the trust around them,” says Yoav Ziv, CEO of BLEND, who will serve as CEO of the merged company, in a joint interview with Erez Moskovitz, founder and CEO of Tasq AI, who will take on the role of president. “Models can write, analyze and draw conclusions, but once an organization doesn’t trust the output, it simply won’t let them make real decisions.” In most scenarios where accuracy truly matters, he adds, a human still needs to remain in the loop, not to replace AI, but to enable it to operate safely.

This trust gap is especially evident in financial and regulated industries. An insurance company may deploy AI models to support underwriting decisions, and an airline may use AI to price tickets, but in both cases handing over full decision-making authority to an algorithm remains difficult. “A single mistake can cost a lot of money,” says Moskovitz. “To adopt models and agents in an enterprise, you need a process of refinement and adaptation to the organization’s data and domain. You don’t just take a foundation model and let it run.”

According to the two executives, the real challenge lies precisely in the transition from a general-purpose model to a specific organizational use case. Training, testing, fine-tuning and then testing again once the model is live in production are essential steps. This is where the merged company’s platform comes into play, enabling a layered combination of automation and different levels of human expertise. Tasks are broken down into small units, filtered automatically, reviewed by broad contributor pools, and only in complex or high-impact cases escalated to domain experts. This approach allows organizations to maintain accuracy and trust even when working with massive data volumes.

The translation world, where BLEND built its business, serves as a natural test case. For years it has been clear that machine translation can handle most of the workload, but without human review, meaning and context can easily be lost. “What happened in translation is now happening across all forms of information processing,” Ziv says. “This is a model that’s being generalized.”

Beyond the technology itself, both executives see the merger as a signal of a deeper shift in the future of work. “We’re moving toward a world where people don’t use AI, but AI uses people,” Moskovitz says. Experts don’t disappear, he argues, but enter the process at critical points, as part of a global network supporting enterprise models and AI agents.

Toward the end of the interview, Moskovitz articulates the company’s ambition in blunt terms. “We want to become Israel’s largest data company.” Behind the statement lies a belief that the next phase of the AI revolution will not be defined by yet another model or chip, but by the ability to refine data, build trust, and turn artificial intelligence into something enterprises can truly rely on at the core of their operations.