This talk was recorded at NDC Sydney in Sydney, Australia. #ndcsydney  #ndcconferences #developer #softwaredeveloper    

Attend the next NDC conference near you: 
https://ndcconferences.com
https://ndcsydney.com/

Subscribe to our YouTube channel and learn every day:   
/   @NDC 

Follow our Social Media!

https://www.facebook.com/ndcconferences
https://twitter.com/NDC_Conferences
https://www.instagram.com/ndc_conferences/

#ai #ethics 

Autonomous AI agents are quickly becoming the next frontier in software development, moving beyond copilots to systems that can sense, reason, and act independently. But with this evolution comes a new set of challenges: agents that operate without clear boundaries, make opaque decisions, or act without human oversight can introduce serious risks to safety, fairness, and accountability.

In this session, we’ll explore how developers can apply the principles of Responsible AI to build autonomous agents that are not only powerful but also trustworthy. We’ll walk through technical implementation strategies using publicly available Microsoft and Azure tools, including orchestration frameworks, fairness assessment toolkits, and governance patterns that enforce safe behaviour.

Through hands-on demos, we’ll showcase practical examples of how to audit agent behaviour, enforce action boundaries, and build systems that align with ethical and regulatory standards. You’ll see how to use tools to embed responsibility into every layer of your agentic architecture.

This talk will equip you with practical techniques to build autonomous AI agents responsibly using open-source tools, design patterns, and governance frameworks that ensure safety, transparency, and accountability from day one.

NDC Conferences

A conference talk covering how to build responsible autonomous AI agents, focusing on safety, fairness, accountability, and security. Key topics include the OWASP Top 10 for agentic applications, agent failure modes (memory poisoning, prompt injection, cross-agent attacks), and a practical demo of a health-tracking agentic system called BioTracker. The talk covers defense-in-depth patterns: tool whitelisting and budgeting via middleware, Microsoft Entra Agent ID for agent identity and zero-trust between agents, independent reviewer agents for output validation, Azure AI Content Safety for groundedness and prompt shields, and OpenTelemetry-based observability. Red teaming with PyRIT and continuous adversarial evaluation in production are also discussed, along with emerging regulations like the EU AI Act.

Building Responsible Autonomous AI Agents: Principles, Patterns, and Practices - Will Velida