Redpanda Data

@redpandadata

Built to handle the world's most critical data, Redpanda now governs the AI agents that act on it.

Global

redpanda.com

Replies Reposts Media

Pinned
Redpanda Data
@redpandadata
Aug 31
You can't govern an AI agent with a prompt it's allowed to ignore. Most governance stops at the application boundary. Agents don't. Every handoff is a governance gap. The Redpanda Agentic Data Plane closes them: one governed layer, every request. See it in action ➡️
00:00
Redpanda Data
@redpandadata
Sep 1
We've just been named a Leader across 10 of @G2dotcom's 2026 reports for Event Stream Processing! 🔥 Huge thank you to every customer who left a review. From what's working to what we should build next, your feedback shapes where we go. 🚀 Read the reviews for yourself:
1
Redpanda Data
@redpandadata
Sep 1
Your board wants AI agents. Your engineering team is quietly explaining that the data isn't ready. On Sept 10 @GlobalFoundries CIO Vishal Mehra joins Redpanda CEO @emaxerrno for a live fireside chat to talk about how he put AI agents into production, what he turned off first,
Redpanda Data
@redpandadata
Aug 31
Most agent guardrails live around the model. CoreBreak showed three major stacks could be made to act without the model running at all: forge a tool call, or forge the human's "approved." Fix: enforcement the agent can't reach. Ours predates the CVEs.
CoreBreak proves agent guardrails need to live outside the agent
From redpanda.com
Redpanda Data
@redpandadata
Aug 31
At Black Hat 2026, Aviyam Ivgi and Hedi Ingber showed the same structural flaw in three of the largest agent stacks in production: AWS Bedrock AgentCore with the Strands SDK, Google's Agent Development Kit, and Vercel's AI SDK harness. They called it CoreBreak. Most AI agent
CoreBreak proves agent guardrails need to live outside the agent
From redpanda.com
1