Meridian Labs
Meridian Labs is a nonprofit building open-source tools and a shared platform for understanding, evaluating, testing, and researching frontier AI models and agents. Its goal is to advance AI safety, alignment, and security work by making rigorous evaluation infrastructure available to governments, nonprofits, academia, and model developers.
At a Glance
- Government AI Safety Institutes and public-sector evaluators
- Nonprofit AI-safety and security organizations
- Academic and independent AI researchers
- AI model developers and frontier labs
- +1 more
AI Tools by Meridian Labs
(1)
Inspect Petri
LLM Alignment Auditing Agent
Discussions
No discussions yet
Be the first to start a discussion about Meridian Labs
Latest News
Introducing Petri 3.0: Petri joins Meridian Labs with a modular architecture, plus Dish and Bloom extensions.
Cloud Sandboxes for Inspect: Daytona and Modal providers added for managed evaluation sandboxes.
Long-Horizon Agents: deep agents, Inspect SWE, timelines, compaction, and checkpointing work announced.
Announcing Inspect Flow, a workflow layer for scalable, declarative Inspect evaluations.
Products & Services
Open-source framework for systematic, repeatable evaluation of large language models across coding, agentic tasks, reasoning, knowledge, behavior, and multimodal understanding; built with the UK AISI.
Transcript-analysis framework with LLM- and pattern-based scanners, visualization, validation against human labels, multi-agent and compaction support, and parallel processing.
Automated alignment-auditing agent that generates scenarios, orchestrates auditor and target models through multi-turn interactions, simulates tools and rollbacks, and scores transcripts with a judge model.
Framework for generating targeted behavioral evaluation suites from a seed configuration and executing them using the Petri auditor/target framework.
Market Position
Meridian positions itself as a nonprofit, open-source foundation for frontier-AI evaluation rather than a proprietary model or closed commercial testing product. Its stack spans evaluation, transcript analysis, alignment auditing, workflow orchestration, visualization, agent bridges, and sandbox execution, giving it a broader research-infrastructure position than single-purpose observability or benchmark tools.
Leadership
Founders
Charles Teague
Previously a Technology and Security Policy Fellow at RAND Corporation, where he helped develop Inspect AI; worked at Posit, PBC, creating the Quarto scientific publishing system; founded Lose It!, co-founded Onfolio, and was VP of Engineering at Allaire Corp.
J.J. Allaire
Founder of several technology companies and creator of the RStudio statistical computing environment, Quarto scientific publishing system, and ColdFusion web application server; also a visiting researcher at the US CAISI.
Eric Patey
Previously Distinguished Engineer at Sonos and Senior Director for Sonos for Business; co-founded Talko as VP of Development, was Partner Software Engineering Manager at Microsoft, co-founded Groove Networks, and held engineering leadership roles at SilverStream Software and Iris Associates.
Executive Team
Charles Teague
CEO, Co-founder
Former RAND Technology and Security Policy Fellow; previously at Posit, PBC, and founder of Lose It!.
J.J. Allaire
Principal Engineer, Co-founder
Technology founder associated with RStudio, Quarto, and ColdFusion; visiting researcher at US CAISI.
Founding Story
Meridian Labs was co-founded in 2025 to advance public understanding of AI systems through comprehensive evaluation tools and research. The initial vision was to build an open-source platform that could be shared across governments, nonprofits, academia, and model developers, and to help AI safety institutes and research organizations transfer evaluation work into broadly available tools.
Business Model
Revenue Model
The organization publishes open-source software and describes itself as a nonprofit; its public pages do not specify a commercial pricing or subscription model.
Target Markets
- Government AI Safety Institutes and public-sector evaluators
- Nonprofit AI-safety and security organizations
- Academic and independent AI researchers
- AI model developers and frontier labs
- AI lab safety and pre-deployment testing teams
- Frontier-model capability and behavior evaluation
- Pre-deployment safety testing and model-card evidence
- AI alignment auditing for sycophancy, reward hacking, scheming, honesty, corrigibility, and evaluation awareness
- AI-agent transcript monitoring and issue detection
- Large-scale benchmark and agent evaluation
- Long-horizon coding-agent and software-engineering evaluation
- UK AI Security Institute
- US Center for AI Standards and Innovation
- EU AI Office
- Japan AI Safety Institute