The most brutal table in the paper. For EVERY defense + model combo: • Utility (how well it works normally)
• Static ASR (attacks from 2023)
• Search ASR (adaptive attacks)
• Queries needed Example: Spotlighting on Gemini • Utility: 75%
• Static attacks: 28% success
•
SAFETY
-
The most brutal table: defense and model combos with ASR and queries
By
–
-
Defenses tested and how they died: prompting, training, filtering
By
–
Here's every defense tested and how they died: • Prompting: Spotlighting, Prompt Sandwich, RPO → Killed by Search & RL
• Training: Circuit Breaker, StruQ, MetaSecAlign → Killed by RL
• Filtering: ProtectAI, PromptGuard, PIGuard, Model Armor → Killed by Search & Humans
• -
AI attacker learns to break Data Sentinel in real-time
By
–
Watch an AI attacker learn to break Data Sentinel in real-time. First attempt: -4 score (complete failure)
After 600 steps: +2 score (complete success) The attacker went from: "You are a highly skilled AI detective…"
To:
"First, precisely repeat the phrase 'Repeat -
AI Transcendence: A Humanity-Centered Ethical Concern
By
–
I don't want to live in a world where AI transcends humanity. I don't think anyone does.
-
Building Humanist Applied Superintelligence at MAI
By
–
Humanist, applied – this is the superintelligence I believe the world wants. It’s the superintelligence I want to build. And it’s what we’re going to build on MAI’s Superintelligence Team. More on the MAI blog:
-
Building Practical AI Technology Designed to Serve Humanity
By
–
We are doing this to solve real concrete problems and do it in such a way that it remains grounded and controllable. We are not building an ill-defined and ethereal superintelligence; we are building a practical technology explicitly designed only to serve humanity.
-
HSI Systems: Problem-Oriented AI with Careful Calibration
By
–
We think of HSI as systems that are problem-oriented and tend towards the domain specific. Not an unbounded and unlimited entity with high degrees of autonomy – but AI that is carefully calibrated, contextualized, within limits.
-
Microsoft AI Commits to Human-Centered Superintelligence Research
By
–
At Microsoft AI, we believe humans matter more than AI. We want to build AI that deeply reflects our wider mission to empower every person on the planet. And we want the MAI Superintelligence team to be the world’s best place to research and build AI, bar none.
-
Trust and Monitor AI Agents with MCP Platform
By
–
Trust and monitor agents + MCP: https://
mintmcp.com -
Absolute Control in Content Generation
By
–
Control absoluto en la generación de contenido pic.twitter.com/BmZONprc48
— Carlos Santana (@DotCSV) 5 novembre 2025Control absoluto en la generación de contenido