Scale partners with Mayo Clinic to develop reliable AI for healthcareRead the Full Story

Scale AI Blog

Matthew Siegel

Insights Generator: Diagnosing Agent Failures

Arabic-First AI: Building From the Ground Up

Agents Can Now Improve Other Agents

Can AI Agents Do the Work of Drug Discovery?

When AI Agents Ask, Attackers Can Answer

Deployment Lessons from Global Governments

SWE Atlas is Complete: Measuring Coding Agents Across the Engineering Loop

HiL-Bench: Your Agent is Smart. It Just Won't Ask for Help.

Understanding New Biosecurity Risks Posed by LLMs

Voice Showdown: The First Arena for Voice AI

Testing & Evals

Testing & Evals

Can Coding Agents Become Engineers? We’re Finding Out.

MoReBench: Evaluating the Process of AI Moral Reasoning

Real Speech Breaks AI (And What We're Doing to Fix It)

Crumbling Under Pressure: PropensityBench Reveals AI’s Weaknesses

The Limits of Data Filtering in Bio-Foundation Models

Breaking Out of the Lab: Testing AI in Professional Domains

The Remote Labor Index: Measuring the Automation of Work

VisualToolBench: Testing the Limits of AI Vision

Enterprise Reinforcement Learning with Rubrics as Rewards

Smoothing Out LLM Variance for Reliable Enterprise Evals

TutorBench: Grading the Next Generation of AI Tutors

Using Rubrics to Build Better Models

AI Doesn’t Live in Text Alone

New Benchmarks Envision the Future of AI in Healthcare

The AI Risk Matrix: Evolving AI Safety and Security for Today

The Future is Multilingual: Scale's New Evaluation Benchmark

WebGuard: A Guardrail for the Agentic Age

Do AI Tools Slow Down Developers? The Answer Isn't Simple.

Beyond the Black Box: Teaching Models to Verbalize Reward Hacking

Detecting and Evaluating Agent Sabotage

I’m Afraid I Can’t Let You Do That

Testing & Evals

Testing & Evals

Pioneering the Era of Experience: Where Human Data Meets Agentic Interaction

The Future of AI Learning Environments: Verifiable Reward + Multi-Agent Interaction

It’s Time to Rethink Red Teaming

LLMs Are Getting Better at Generating Short Fiction

Using LLMs While Preserving Your Voice