Date: June 17, 2026
Time: 9:30 AM EDT | 8:30 AM CDT | 7:00 PM IST
Topic: Data Poisoning Attacks Against Machine Learning Models & Agentic AI Systems
Abstract:
Machine learning models are only as trustworthy as the data they learn from, but that underlying data is increasingly becoming a target. Data poisoning attacks subtly manipulate training datasets, embedding malicious patterns that can degrade model performance, introduce bias, or create hidden backdoors. As organizations accelerate AI adoption across cloud environments, these risks often go undetected until real-world damage occurs.
Unlike traditional cyberattacks, data poisoning operates silently within the AI lifecycle, bypassing perimeter defenses and evading conventional security tools. A single compromised dataset can ripple across multiple models, leading to flawed predictions, regulatory exposure, and erosion of trust. From recommendation systems to fraud detection and autonomous systems, the consequences can be severe and difficult to trace back to the root cause.
The current webinar will demystify data poisoning attacks and equip you with practical strategies to defend against them. We’ll explore how attackers exploit data pipelines, demonstrate real-world attack scenarios, and outline robust mitigation techniques, from secure data sourcing and validation to AI-aware monitoring and governance frameworks. Attendees will walk away with actionable insights to strengthen the integrity and resilience of their machine learning systems.
Key Takeaways:
- Understanding how data poisoning attacks work
- Identifying vulnerabilities across the ML lifecycle
- Explore real-world examples and their business impact across industries
- Leveraging AI security frameworks and governance models to reduce risk and ensure compliance
- Integrating continuous monitoring and logging to detect and respond to attacks
- Aligning AI security across various environments for end-to-end protection
Speaker:
Jason Ross, Product Security Principal at Salesforce
Bio: Jason Ross is a cybersecurity professional with 20+ years of experience, currently serving as Product Security Principal at Salesforce. His work focuses on adversarial testing and defense of deployed generative AI applications, agentic systems, and the large language models powering them. He specializes in prompt injection attacks and defense, model governance and security, and agent exploitation across high-stakes, high-visibility production deployments.
Beyond his role at Salesforce, Jason co-leads the OWASP GenAI Security Project Red Team Initiative and contributed to authoring the OWASP GenAI Red Teaming Guide. He is also an active contributor to the broader security community, serving as a staff member at BSidesLV and a volunteer at DEF CON.
