Building deterministic security for multi-agent AI workflows | David Gildea (Druva)
Автор: Credal AI
Загружено: 2025-09-11
Просмотров: 67
Описание:
David Gildea, VP of Product for AI at Druva, discusses why traditional security frameworks fail in agentic AI environments and how his team is building deterministic security harnesses for their AI products. He touches on why MCP adoption outpaced A2A specifications, how to design AI layers that work better than traditional APIs, and why hyper-personalization will transform enterprise software. David also shares practical insights on building enterprise AI systems that maintain security while enabling autonomous agent workflows.
Chapters:
0:00 Introduction
0:24 David's background at Druva and AI product adoption
2:04 Real vs perceived AI security threats
3:32 MCP vs A2A specifications comparison
6:02 Missing pieces in MCP authentication
8:57 Agent identity management challenges
12:03 Human-in-the-loop scaling problems
14:21 Supervisor agents for security decisions
17:10 AI layers vs traditional APIs
20:06 Enterprise MCP server documentation challenges
23:46 Building objective-based AI experiences
25:54 Lessons from building agentic systems
29:11 AGI timeline predictions
32:22 Future of AI interaction models
37:02 Druva's new agent framework
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке:
