AI/GenAI/Agentic AI Summit @Check Point

In this talk, Prof. Balicer will describe how modern healthcare is transforming with AI as a key catalysation agent, accounting for opportunities as well as risks in this process.

Add to Calendarcalendarcalendarcalendarcalendar

Understanding and accepting decisions made by artificial intelligence (AI) systems is crucial for human collaboration and trust. This importance grows even further in multi-agent environments, where AI systems make decisions based on unknown goals, potentially influenced by the preferences of other agents. In such complex scenarios, explanations become important in increasing user satisfaction and acceptance by considering factors like the system's decision, user preferences, agent preferences, environmental settings, and key attributes such as fairness, envy, and privacy. In this talk, we will explore the concept of Explainable Decisions in Multi-Agent Environments (xMASE) through three cases: explanations of constraints-driven optimization problems taking into account privacy consideration, explaining decisions of agents in mixed-motive games, advice explanations in complex repeated decision-making environments, and explanations for multi-agent Reinforcement Learning (RL). For each case, we propose an algorithm that generates comprehensive explanations. Furthermore, we will report on human experiments that demonstrate the informativeness and acceptability of explanations to users.

Add to Calendarcalendarcalendarcalendarcalendar

This study explores the detection and analysis of militant Jihadist preaching on social media, where violent incitement is often framed as religious messaging. By examining 6000 posts from Darknet forums associated with Jihadist groups (2017–2018), we apply Reciprocal Human-Machine Learning (RHML) to improve the identification and understanding of such content. RHML maintains a “human in the loop” approach, allowing human input to mitigate machine bias and prevent performance decline while fostering mutual learning between humans and machines.

To support this, we adopt a design science methodology to create a theoretical framework for RHML, particularly in the context of text classification tasks. Our findings demonstrate that RHML not only enhances detection capabilities but also provides deeper insights into the influence tactics used by terrorist groups, thus advancing counterterrorism efforts through improved understanding and identification of militant messaging.

Add to Calendarcalendarcalendarcalendarcalendar

Most real business value doesn’t come from generating content - it comes from anticipating outcomes and acting on them. Predictive agents that turn natural language into structured, executable predictions represent a more meaningful evolution of enterprise AI. These systems don’t produce text. They produce decisions. I’ll outline their core architecture, the challenge of mapping conversation to predictive logic, and what it takes to abstract data prep, target definition, and modeling without compromising reliability.
My claim is this: in business, the real Turing test isn’t fluency. It’s accuracy, autonomy, and the ability to be trusted with consequences.

Add to Calendarcalendarcalendarcalendarcalendar

AI is crossing a threshold — from chatbots to autonomous agents, from closed models to open ecosystems. This keynote explores the breakthroughs fueling this transformation, the interoperability enabling collaboration, and the new mindset required to lead and secure the AI battlefield of tomorrow.

Add to Calendarcalendarcalendarcalendarcalendar

This talk reveals how AI Multi-Agent Systems (MAS) are transforming cybersecurity through orchestrated intelligence rather than isolated tools. Drawing from Twine Security's enterprise implementations, this session will demonstrate how specialized AI Agents work in concert and can manage Identity workflows, detect threats, and remediate vulnerabilities with maximum efficiency. The presentation offers a compelling vision where human cybersecurity teams collaborate with AI Digital Employees, proving why this approach is the future of the cybersecurity workforce.

Add to Calendarcalendarcalendarcalendarcalendar

In this presentation, we uncover the groundbreaking yet unsettling role of AI in reshaping the battlefield of cyber warfare. As AI evolves at breakneck speed, it’s not just revolutionizing industries—it’s empowering hackers to unleash advanced cyber weapons at unprecedented efficiency. These tools drastically cut down the time required for attacks, creating a new era of high-speed automated cybercrime.

Add to Calendarcalendarcalendarcalendarcalendar

The recent GenAI-based applications by Maccabi Healthcare Services, now integrated into various physician systems, designed to optimize workflow and improve the precision of medical care.

"Bat Kol", a GenAI-powered platform that transcribes doctor-patient encounters and generates clinical summaries and integrating them into the physician’s operational systems—saving valuable time and reducing documentation burden.

Another application is a GenAI-based clinical assistant that provides physicians with real-time, chat-based access to thousands of clinical protocols directly within Maccabi’s systems, enhancing decision-making at the point of care.

Add to Calendarcalendarcalendarcalendarcalendar

This lecture offers a practical deep dive into designing intelligent LLM agents for real-world deployment in high-stakes domains such as finance and cybersecurity. We will examine the transition from static workflows to dynamic, agentic systems, emphasizing the core capabilities that define agent behavior - and explore strategies for balancing control and flexibility to achieve both scalability and stability. The session will cover how to extend standard RAG pipelines using advanced NLP and GenAI techniques, integrate modular workflows, and make informed tradeoffs between closed-source APIs and private open-source models - including ways to effectively combine the two. Finally, we’ll discuss structured evaluation strategies that go beyond accuracy, such as step-level critique, regression safeguards, and trust-oriented testing. This talk is grounded in implementation best practices and real-world lessons for teams building secure, adaptive, and high-performing LLM agents.

Add to Calendarcalendarcalendarcalendarcalendar