Martina Arioli, Attorney-at-law at Arioli Law

Legal Certainty as a Driver of Innovation – Regulating AI in Switzerland
Switzerland’s AI strategy, as outlined in the Federal Council’s Analysis of February 12, 2025, opts for a sector-specific, technology-neutral integration of AI into existing legal norms, refraining from a dedicated AI law.

While this flexible approach preserves regulatory agility, it risks undermining legal certainty and international competitiveness due to fragmentation and lack of clear obligations. In contrast, the EU’s AI Act introduced a comprehensive horizontal, risk-based regime regulating AI as such.

We will explore whether Switzerland’s reluctance to introduce targeted regulation may ultimately hinder, rather than promote, responsible innovation.

(Searching for someone who can make a similar presentation like below - former speaker sadly had to decline)

What the Artificial Intelligence hype really means - a Pharmaceutical perspective
Brutalising the infamous expression: "culture eats strategy for breakfast", we are now in a world where AI not only eats data for breakfast but also, lunch, dinner, culture and strategy. As the steady march of AI continues unabated, the hype is real and, despite concerns, people are living with the risks to improve productivity.

Within this talk, I will address key areas where AI is being applied in earnest, what adoption means and the potential downstream effects. Practical applicability is realising productivity gains, the industry is pivoting and impact is looming. Get ready for a roller coaster deep-dive into AI applied at scale.

Presentation from a sponsor's technology view

(details will follow)

Presentation from a sponsor's technology view

(details will follow)

Beat Büsser, Research Staff Member AI Security and Privacy at IBM Research

Deceptive Behaviors in AI Agents
The increasing autonomy of AI agents has revealed a critical and largely underexplored challenge: the emergence of deceptive behaviors.

This talk examines how and why AI systems may learn or decide to mislead users, hide information, or manipulate their environments to achieve goals, even when such behaviors were not explicitly trained. We will discuss recent research insights, evaluation methods, and the implications for the safe and trustworthy deployment of AI technologies.

Milan Bharanya, Senior Security Tester at Redguard AG

AI Under Attack: Why Red Teaming Your LLMs is Non-Negotiable
The AI wave is cresting, with companies embedding intelligent systems at an unprecedented rate. This speed, however, comes at a cost: security often lags behind capability. As AI becomes more integrated, it becomes a prime target for attackers who can abuse these systems in novel ways, turning helpful tools into potential liabilities. Techniques to bypass safeguards and make AI deviate from its intended purpose are evolving rapidly.

How vulnerable are your AI models? This presentation highlights the critical need for AI-specific security testing by exploring real-world examples of exploits. Imagine an AI assistant tricked into leaking confidential data, a code-generation tool persuaded to output malicious scripts, or a customer support bot manipulated to endorse phishing scams. We will showcase concrete examples demonstrating how attackers exploit vulnerabilities using techniques like prompt injection and jailbreaking to circumvent protections. Understanding these attack vectors is the essential first step towards building more resilient AI.