Alice x Amazon Webinar: Navigating Agentic AI Risks With Agentic Solutions
Join an exclusive briefing on the future risks, and safety opportunities, of AI agents.
Sign Up Now
Watch On-Demand
Alice x Amazon Webinar: Navigating Agentic AI Risks With Agentic Solutions
Overview
AI systems are rapidly evolving - not just in capability, but in autonomy, agency, and interconnectivity. As these systems begin to reason, plan, and act with increasing independence, the associated risk landscape is shifting as well
In this exclusive executive briefing, Iftach Orr (CTO, Alice) and Charith Peris (Senior Applied Scientist, Amazon AGI) will delve into the next generation of AI risks - focusing on vulnerabilities already surfacing in real systems and what it means for those designing and deploying foundation models.
A central theme: agentic AI. These systems go beyond passive generation. They take actions, invoke tools, collaborate with other agents, and operate in dynamic environments. These capabilities create new forms of emergent risk - but they also unlock new ways to improve safety, alignment, and control.
To illustrate, Charith will share Amazon’s recent work on using multi-agent deliberation to generate high-quality chain-of-thought (CoT) training data - showing how agentic systems can be used to enhance policy adherence, jailbreak robustness, and overall model safety.
What You’ll Learn
- The most urgent and emerging risks in Agentic AI safety and security
- Real-world examples of vulnerabilities, including malicious code generation, agentic protocols exploitation, data exfiltration, scaled fraud, reasoning injection and more
- How agentic AI can also improve safety – including a look at Amazon’s multiagent framework for generating safer chain-of-thought data
- Practical steps companies can take now to future-proof their Agentic AI
Meet our speakers
What’s New from Alice
Introducing Guardrails Trained for Your Policies
Generic guardrails weren't built for your policies. WonderFence trains a custom detector for each one, using adversarial data from years of protecting the world's largest tech platforms, so you can deploy consumer-facing AI without compromise.
What Does It Actually Take to Build Unbiased AI?
Nobody told Tennisha Martin the importance of having a mentor, so she built a community of tens of thousands instead. As the Founder and Chairwoman of BlackGirlsHack, her whole mission has been making sure nobody else has to figure it out alone. In this episode, she and Mo get into AI bias, why it's already showing up in places that matter far beyond tech, and why the real fix starts with getting the right people in the room when these systems get built.
Distilling LLMs into Efficient Transformers for Real-World AI
This technical webinar explores how we distilled the world knowledge of a large language model into a compact, high-performing transformer—balancing safety, latency, and scale. Learn how we combine LLM-based annotations and weight distillation to power real-world AI safety.
Building AI Applications in Financial Services
A practical guide to building safe, compliant AI applications in financial services, covering governance, model risk, and regulatory obligations across the full development lifecycle.