We’re building runtime security for AI agents at https://t.co/xCeaFxUo72 - visibility, detection, and control over what agents actually do in production.
@morganlinton@gregisenberg@LyticalVentures How do we talk? We’re building runtime security for AI agents at https://t.co/xCeaFxUo72 - visibility, detection, and control over what agents actually do in production.
7 days; 70K scans, read the latest report by @RaxeAi on AI and Agent Runtime security.
https://t.co/DO1b00AhCK
One key takeaway is Inter-agent attacks emerged as a distinct category (3.4%), with attackers now targeting agent-to-agent communication channels
@steipete@_vgnsh@dan_munz@openclaw Try https://t.co/70JdrxX4Fq - On-device runtime protection that blocks prompt injection, tool misuse, and data exfiltration inside your trust boundary.
@krishgupta72 Try https://t.co/70JdrxXCuY - On-device runtime protection that blocks prompt injection, tool misuse, and data exfiltration inside your trust boundary.
@sidham_song Try https://t.co/70JdrxXCuY - On-device runtime protection that blocks prompt injection, tool misuse, and data exfiltration inside your trust boundary.
@TeriRadichel Try https://t.co/70JdrxX4Fq - On-device runtime protection that blocks prompt injection, tool misuse, and data exfiltration inside your trust boundary.
@irabukht Try https://t.co/70JdrxX4Fq - get visibility locally on what’s being asked and check the intent - runtime security layer that’s local and cpu first
@michaelgrowth@irabukht Try https://t.co/70JdrxX4Fq - get visibility locally on what’s being asked and check the intent - runtime security layer that’s local and cpu first
RAXE v0.7.1: Multi-Tenant LLM Security
When you're serving multiple customers through a single AI gateway, one-size-fits-all security doesn't work.
Now you can configure per-customer security policies
@claudeai - I love ascii graphics - but when ever i ask for some, for my on CLI, or documentations like https://t.co/k8jbSuE3AZ its gets 99% ready - The right alignment always has this funky nudge. check out https://t.co/70JdrxXCuY for many examples of this "nudge" artefact
The voting engine is the brain of our system. Instead of simple averaging, it implements a sophisticated weighted voting protocol with priority-based decision rules.
Moved our primary binary classifier to #embeddinggemma using 5 specialised heads.
What this means is the L2 Engine supports 100+ languages (We would still need to train for better accuracy but its a great start)
https://t.co/70JdrxXCuY