Palo Alto CEO Warns: Current AI Agent Tech Stack Has Zero Security

brucemacv · x · 2026-07-20

Nikesh Arora, CEO of Palo Alto Networks, pointed out that the current enterprise AI Agent tech stack has severe security vulnerabilities.

Lack of Authentication: There is currently no secure version of the MCP (Model Context Protocol). Agents lack identity verification when interacting with each other or platforms like Salesforce.
Missing Asset Inventory: Enterprises generally have no idea how many models are deployed internally. Millions of models downloaded daily from platforms like Hugging Face enter corporate environments directly.
Supply Chain Risks: These unvetted models might contain backdoors or dormant malicious connections, a fact that companies remain completely oblivious to.

More from coding & agent