From Provenance to Continuous Verification
Cisco’s AI Supply Chain Provenance Explorer strengthens the evidence available before deployment. Model Informatics extends verification into runtime.
Project VAIL
VAIL's objective is to create a universally accepted standard for confirming the legitimacy of AI/ML algorithms, comparable to the 🔒 visible in a browser tab (akin to the SSL/TLS on the internet).
Live Last read · last published · next check
Latest posts
Measuring the Integrity of Infrastructure for Frontier Intelligence
What happens to older Claude models when a newer one ships and why "pin the version" isn't the safety net you think it is.
Misbehaving Agents & The Impacts of Extreme Instability
Endpoints with extreme levels of behavioral instability show high levels of task instability compared to peer endpoints serving the same nominal model.
Introducing Stability Arena and The Seven Metrics of Model Hosting
Why Latency, Throughput, Error Rate, and Cost Are No Longer Enough
Reliability ≠ Stability
We shouldn't assume reliable model APIs mean we are getting intelligent outputs.
Building a Vocabulary for AI Assurance
Part II: Establishing Ground Truth via Verifiability and Accuracy
Building a Vocabulary for AI Assurance
Part I: Making Sense of Explainability and Interpretability
What is Model Informatics? What does it mean to "Verify" AI?
We first started exploring verifiability almost 3 years ago and first shared a few initial ideas in early 2024.
Agents, Task Time Compute, & Task Time Marketplaces
Agents coordinating to complete complex, multi-step tasks for users and using a marketplace to bid out individual tasks to specialty agents.
The Security Evolution of Core Technologies: What It Means for AI (Part I)
There is a consistent pattern of increasing the robustness of security features and capabilities for core computing technologies. AI shouldn’t be any different.