Sarine Technologies Ltd.

M&A Activity • Aug 18, 2025

ELEASEבין

PRESS RELEASE

Sarine Acquires a Minority Stake in Kitov.Ai

Hod Hasharon (Israel), 18 August 2025 – Singapore Exchange Mainboard and Tel Aviv Exchange listed Sarine Technologies Ltd ("Sarine" and along with its subsidiaries "the Group") (U77:SI; SARN.TA), a worldwide leader in the development, marketing and sale of precision technology solutions for the evaluation, planning, processing, measurement, grading and trading of diamonds and gems, further to its announcement of 23 February 2025 regarding the execution of a Letter of Intent (LOI) in this matter, wishes to update the public on its closing of its acquisition of a minority stake in Kitov.ai.

The purpose of this investment is the diversification of Sarine's focus to additional industries, also considering the current challenges the diamond jewellery industry faces. Being a company engaged in technologies similar to those employed by Sarine (optical inspection, AI, software, etc.), thus "speaking" the same language, Kitov.ai provides the Group with the means to diversify into new fields separate from the diamond industry. Kitov.ai has many industry-leading customers in many varied industries including aerospace, defense, electronics, medical devices, energy control, consumer products and others, in the U.S., Europe, the Far East and Israel.

The concluded deal includes an initial cash investment of US\$ 4.1 million in consideration of a just over 33% stake in Kitov.ai, paid in part to the existing shareholders of Kitov.ai and in part infused into Kitov.ai as working capital. Sarine is also lending Kitov.ai an additional US\$ 2.6 million, in the form of a convertible loan, which, not before 01 January 2027 and not after 15 February 2028, can be converted, at Sarine's sole discretion, into additional equity shares, bringing Sarine's total stake in Kitov.ai to 51%. The transaction was conducted based on Kitov.ai's pre-money valuation of US\$ 10.635 million (such valuation was set as a result of arms-length negotiations between the parties, based on a business plan presented by Kitov.ai and reviewed by the Company)

If Sarine does decide to convert into equity said convertible loan, possibly subject to shareholders' approval, the following arrangements shall apply:

• Sarine may be required to pay (during 2027 or during the first quarter of 2028) certain current shareholders a conditional earnout of US\$ 1.0 million triggered by meeting defined profitability goals;
• Kitov.ai, the company itself, not Sarine, may be required to pay (in 2029) certain current shareholders an additional conditional earnout payment of up to US\$ 2.0 million, triggered and calculated by meeting certain sales and profitability goals;
• The current shareholders of Kitov.ai may exercise (during the second half of 2029) a Put Option and sell to the Company all or part of Kitov.ai shares held by them at such time, based on a valuation set by an independent appraiser, at a consideration not to exceed

US\$12 million. This Put Option will not be in force, if before mid-2029 equity altering events occur, such as a merger, a public offering, etc.

If Sarine does not convert the the aforesaid convertible loan, the aggregate consideration payable by Sarine for its 33% of Kitov.ai's shares shall be ~US\$ 4.1M and the said convertible loan shall bear interest and be payable to Sarine.

The investment documents also include minority-protection clauses, throughout the different phases of the above transaction.

Kitov.ai has developed and markets an AI‑driven automated 3D visual inspection system that can concurrently check products for the existence of key elements and their correct positioning along with the inspection of surface finishing, labels, barcodes, etc. Kitov.ai's solution can be implemented, due to its flexible software characteristics, to inspect completely different parts at the same installation point in high‑mix/low‑volume manufacturing environments.

Kitov.ai's patented CAD2Scan AI-based inspection definition software package is unique in that the design engineer utilises the part's CAD model to directly define the inspection requirements (where to inspect, what to inspect for, pass/fail criteria, etc.) in minutes, without considering computer vision or automation issues, and without having to "walk" the robot through the inspection steps, creating a flexible inspection regime, reducing engineering effort and accelerating new‑product conversion from prototype to manufacturing.

Kitov.ai's patented technologies automatically select the best viewpoints and lighting1 for each feature inspected and compute the most efficient robot path between them – optimising the eyes (vision), the hand (robotics) and the brain (AI) integration and improving failure detection beyond conventional approaches. By integrating intelligent robotic image acquisition, CAD-based part definition2 , cross-part learning, as detailed below, and even multi-sensor feedback (Kitov supports

² Beyond using the CAD model to plan camera motion, Kitov feeds 3D structural information into the detection process itself. By leveraging the CAD data, Kitov's AI is armed with additional features and context that improve its accuracy. The AI is provided with not just raw pixel data, but also information like "surface normal vectors" or curvature at each location and the expected appearance of the inspected feature, if it were perfect. which helps distinguish actual defects from benign texture or lighting effects. A subtle shadow that confuses a usual inspection algorithm is recognised as just a curvature change. Similarly, if the CAD file specifies a certain feature, say a label, the system won't flag it as a foreign object or damage. This kind of 3D-awareness enhanced inspection is unique to Kitov's hybrid approach, which combines classic 3D computer vision with modern AI. The outcome is a more powerful inspection paradigm. Kitov's exclusive use of the CAD data bridges the gap between how engineers use their drawings to define the necessary inspection and how the AI uses the same data to become better at defect detection

4 Haharash St., Neve Neeman Hod Hasharon, Israel 4524075 Tel. +972-9-7903500 www.sarine.com

¹ In the field of machine vision, the correct lighting is 90% of the solution- the system's performance hinges on optimised lighting and inspection intensity/angles. If illumination and viewing angles are poorly chosen, the most sophisticated AI will struggle. In essence, what you see is what you can inspect. Kitov AI's first breakthrough is active control of the image capturing. Uniquely using the CAD model its source, Kitov's Smart Planner software computes the optimal 3D positions and orientations from which to view each inspected feature of the object at precisely the right angle and with the right illumination. The planner's AI inherently understands that different surface materials and shapes require different lighting strategies – for example, a shiny curved surface might need a dark-field glancing light to highlight scratches, whereas a matte area might need direct lighting. Kitov maximises the inherent detectability of defects before the follow-on AI even processes the images. Kitov's AI provides optimised robot path planning and defect-enhancing lighting to find every defect, every time.

multiple sensors data acquisition on a single path of the robotic inspection head), Kitov's system pushes far beyond the limitations of conventional machine vision inspection systems.

Another key innovative aspect of Kitov's solution is cross part learning- namely how to overcome the problem of data scarcity, that plagues industrial deep learning applications. In high-quality manufacturing it's common to have a relatively low number of defective samples for any given part type. Though obviously a positive for the manufacturer, training an AI model on only a handful of examples is challenging. Kitov addresses this by utilising images from different parts that have visually similar characteristics and creating a unified defect detection network on the aggregated data. In practice, this means that Kitov's software pools the information derived from the inspection of distinct parts of the same material or finish. A scratch on an anodised aluminum aircraft bracket looks similar to a scratch on an aluminum valve's body. This unified "learning" creates a model of scratches per se, across all similar-looking parts, resulting in a more robust AI model. The AI network detects defects better because it has seen multiple contexts across different shapes and sizes. Kitov's AI gains broader experience, similar to how a seasoned human inspector applies his experience to new products. Kitov's AI breaks the glass ceiling of traditional deep learning, which typically requires training each model on one specific part with lots of examples.

Kitov.ai's open, vendor‑neutral platform (use whichever CAD software, whichever robot, whichever lighting, whichever camera, etc.) integrates seamlessly with existing factory automation and inspection tools, enabling OEMs and integrators to collaborate and quickly deploy solutions within a constantly expanding partner ecosystem.

Kitov.ai's next innovation will be the introduction of its model-based enterprise (MBE) solution in 2026. MBE's benefits for an organisation are that it leverages 3D digital models to integrate and manage technical processes throughout the product's lifecycle. An MBE uses an authoritative 3D model that contains all the necessary information pertaining to the design, manufacturing, inspection, cost and pricing of a product. This enhances collaboration and communication across plant departments, reduces errors and optimises corrective cycles, ultimately leading to faster product development and reduced overall costs. There are two prerequisites to implementing an MBE:

• The first is the creation of annotated 3D models, known as Model-based definitions (MBD). This requires the use of a CAD system capable of creating precise solid models, with product and manufacturing information (PMI). The CAD systems used in aerospace, defense and automotive industries generally have these capabilities.
• The second prerequisite is integrating the MBDs into a unified platform, so they can be used to substantially optimise downstream lifecycle activities. CAD models are typically generated in proprietary data formats and must be translated to a suitable MBD-compatible standard format, such as ANSI QIF. Kitov.ai's next offering will support relevant CAD systems and QIF formats and thus facilitate utilisation of the benefits of MBE for its customers, for instance by feeding AI-identified repetitive flaw data, derived from Kitov inspections, directly into the CAD of redesign/upgrade cycles.

Kitov has filed 22 patents in the United States, Europe, Japan, China and Israel, across 6 patent families, of which 8 have already been granted.

Kitov.ai's core business strategy has been redirected from selling complete tailored solutions to each customer, including robots and other incidental equipment, to focusing on Kitov.ai's core competencies and creating a software-license centered recurring revenue model. As noted above, Kitov.ai's open system architecture enables its customers to easily integrate the system with any leading CAD package or any robotic arm, foregoing the need for Kitov.ai to perform lengthy integration, installation and support processes. Cooperation with other leading companies in the

QA/QC field, by offering solutions based on conjoining other products' benefits, is a sales channel being aggressively evaluated. It is our aim that the new entity be profitable in 2026.

Kitov.ai will retain the founding visionaries and the core R&D team, has engaged a new manager of its U.S. operations and will be engaging new C-suite executives and a new Product/Business Manager. Kitov.ai's founder is Dr. Joseph (Yossi) Rubner, a graduate of Israel's prestigious Technion in Computer Science (Summa Cum Laude). He then went on to earn PhD degrees in Computer Science and Electronic Engineering at Stanford University in California and is an internationally recognised expert in machine vision and artificial intelligence. His company RTC Vision, the parent from which Kitov.ai was spun out, executed the vision and image processing development work on Sarine's Galaxy® systems and our initial Clarity grading system. We are confident our long-established and mutually respectful and beneficial relationship will contribute to the initiative's success.

Mr. Daniel Glinert, Executive Chairman of the Group, said, "I am very excited about this opportunity. Kitov.ai has top-notch technology in an expansive field and an impressive list of customers across a multitude of disciplines. I am sure that with our contribution to their business strategy and execution, I believe we can help Kitov.ai evolve into a leading world-class provider of QA/QC solutions. No less importantly, the opportunity to diversify into other fields comes at an opportune time and at a reasonable price. I wish us all a successful productive future."

About Sarine Technologies

Established in 1988, Sarine Technologies Ltd. is a global leader in developing advanced technologies for modeling, analysis, evaluation, planning, processing, finishing, grading and trading of diamonds. In recent years, Sarine's business has pivoted to deriving mostly recurring revenues from its proprietary Gal3D inclusion and tension mapping (which processes the Galaxy® platforms' output) and Advisor® rough diamond planning cloud-based software packages, along with its other various pay-per-use services. At the heart of Sarine's ecosystem is the Advisor® software, which integrates internal inclusion scanning data and geometrical 3D analyses, to provide rough diamond planning and processing. Sarine's Most Valuable Plan™ (MVP) software, launched in 2024, builds on Advisor® 8.0's capabilities and to provide not only the most advanced but also predominantly automated planning for natural rough diamonds, delivering greater value and enhanced production efficiencies in terms of both time and cost. Sarine's broad array of services, based on data derived from its cutting-edge technologies, also includes Lab-Grown Diamond (LGD) planning, Journey™ provenance and traceability reports, GCAL diamond grading reports and other retail-focused solutions like visualisation and fingerprinting. Sarine continues to develop and sell its worldrenowned products, including the Galaxy® family of inclusion and tension mapping systems, rough diamond modelling platforms, laser-marking, inscription and fingerprinting equipment, automated (AI-derived) Clarity, Color, Cut and light performance grading systems and the AutoScan™ Plus for natural rough diamond source registration. For more information about Sarine and its products and services, visit http://www.sarine.com.

Contact APAC IR

Cyrus Capital Consulting Mr. Lee Teong Sang Tel:+65-96339035 teongsang@cyrus.com.sg Sarine Technologies Ltd. Marketing & Communications Ms. Romy Gakh-Baram Tel:+972-9-7903500 romy.gakh-baram@sarine.com

4 Haharash St., Neve Neeman Hod Hasharon, Israel 4524075 Tel. +972-9-7903500 www.sarine.com