So elegant. Reminds me how powerful backprop & autograd are. We can keep building new gadgety puzzle pieces, use any mix of masking regimes we want & just blend it all together
Multi-objective optimisation was beaten by blessings of dimensionality + automatic differentiation
Investigating suspicious AI workflows in Microsoft Entra Agent ID: Autonomous agents, by @mattifestation
Part 1: https://t.co/X7JO1HvrZB
Part 2: https://t.co/E7FQmztOJm
Tomorrow: static NHI secrets meet runtime identity.
See how SPIFFE and SPIRE issue short-lived workload credentials for agents and ephemeral cloud workloads.
📅 Jun 8 | 11AM ET
🔗 https://t.co/4WOLia5LzD
#NHI#IdentitySecurity#CloudSecurity
Due to GitHub breach, we released a feed for malicious VS Code extensions, auto-generated with IoCs out of Agent Mesh. I hope it's useful. If it is, let me know and we'll do this regularly, Can add in skills, MCP servers, etc.:
https://t.co/vvvGDrCKgj
Let me know if useful.
The scanner reads Office files as ZIP archives in memory and searches PDFs across both raw bytes and Flate/deflate-decompressed streams, filtering common schema domains to cut false positives.
https://t.co/YDi7KtHNrV
On this week’s livestream, we’ll show you how to build a GPU-accelerated multi-agent app: orchestrate a team of agents using ADK, serve Gemma 4 on Cloud Run using NVIDIA RTX PRO 6000 GPUs, and connect them with Milvus.
Tune in on May 12 at 9 AM PT → https://t.co/PKhK9eN7sE
🚨BREAKING : Call centers are officially dead.
ElevenLabs Agents quietly wiped out the $40B customer support industry.
→ Sounds human in 70+ languages
→ Books, updates, closes tickets mid-call
→ Plugs into GPT, Claude, Gemini, any LLM
→ $0.08/min, startups get $4K free
Revolut, Cisco, Deliveroo already switched.
You're next 🧵
MFA isn't enough. Trusted updates became weapons. And AI agents are acting outside expected behavior.
Ismael Valenzuela is connecting the dots on The Replicant Problem LIVE at #SecureYourFortress 🤖🔍
Tune in → https://t.co/lIrrQw1fLw
#ZeroTrust#AIAgents#CyberDefense
Anyone can join GEAR for access to AI training and more, but if you're at #GoogleCloudNext, we've also integrated this new program powered by Google Skills throughout the event—making it easy to build skills on site and jumpstart your GEAR journey → https://t.co/xg7yi4J6WV
You know Wireshark's GUI — but what about its CLI tools? Andy Laman explores the command-line side of Wireshark at #SecureYourFortress — free & virtual, Apr 30. 👉 https://t.co/4NH3rLFDaK
FREE Azure Red Team Webinars
As part of the Month of Azure Red Teaming 2026, explore real-world attack techniques across compute, identity, APIs & more.
Learn from experts through hands-on sessions.
Register:
https://t.co/Mcut8W6fzh
#Azure #RedTeaming #CloudSecurity #CyberSecurity
Claude Code's remote control protocol lets developers orchestrate instances programmatically. @tyholms reverse engineered it and found an undocumented flag that redirects any instance to attacker-controlled infrastructure, silently bypassing all permission checks.
https://t.co/Tn85uj1R77
If you can't see the queries, you can't secure the data. DNS filtering helps you identify and score every AI tool on your network.
See 5 ways to build a better security foundation: https://t.co/hlkWTfH2La
The Initial Access Broker market is maturing. In H2 2025...
📈 Asking prices (and the size of targeted orgs) rose
👀 New marketplaces thrive; older forums stall or shut down
🏛️ Government the top-targeted sector
For key findings, recommendations & more: https://t.co/Sz7UMq2BUb
#TeamPCP was caught using #AdaptixC2 after compromising a package. They deployed the beacon for remote shell, process management, and screenshot capture.
Full writeup by OX Security:
https://t.co/cDu4qx4m6E
The sample hit both of our rules.
https://t.co/T9ZfeH84Iy