Our Investment in mpathic

AI is increasingly the first thing people talk to in their hardest moments. A teenager spiraling at 2 am. A patient parsing a frightening diagnosis. A person in crisis reaching out to the only thing available 24/7. These interactions are happening right now at a massive scale, and the AI on the other end is ill-equipped to safely meet its users’ needs.

This is not a hypothetical risk. As foundation models become the default interface for mental health support, medical triage, and youth engagement, the gap between what AI can say and what it should say is widening fast. The companies building these models know it. They’re actively seeking partners who can bring clinical rigor to a problem that engineering alone cannot solve.

Enter, mpathic.

mpathic is a clinician-founded AI safety company that helps foundation model developers and LLM-powered application teams stress-test, evaluate, and monitor AI behavior in high-risk scenarios. Think “clinical-grade QA for AI” — or, in Foundry parlance, safety “glue” for the AI stack.

The company’s approach is rooted in something most AI safety efforts lack: real clinical expertise. mpathic has built a global network of thousands of licensed mental health clinicians who create ground-truth evaluation datasets, identify failure patterns in sensitive conversations, and monitor live interactions with safeguards that can flag, redirect, or intervene when needed. Grin Lord, mpathic’s CEO, describes it well: “It’s kind of similar to people that create synthetic data for visual AI. It’s not every day a child runs in front of a Waymo, but we can simulate it 10,000 ways with synthetic data. That’s basically what we’re doing, but from a psychological angle with language.”

The results speak for themselves. In one early engagement with a foundation model builder, mpathic’s clinician-led evaluation program helped reduce undesired and dangerous model responses by more than 70%. The company is working with leading AI model developers serving tens of millions of users, alongside clinical partners including Seattle Children’s Hospital and Panasonic WELL.

mpathic is led by co-founders Dr. Grin Lord and Dr. Danielle Schlosser (CIO). Based in Seattle, Grin is a board-certified psychologist and NLP researcher who saw children, clinicians, and patients increasingly turning to AI tools before human support — often in moments of real distress. Together with Danielle, a leading Bay Area clinician and tech innovator, they’ve built the team that has become the foremost experts in translating real-world clinical judgment into repeatable safety workflows for AI systems. They’re a mission-driven team tackling one of the most consequential problems in AI, backed by deep domain expertise and the credibility to match.

To fuel mpathic’s expansion, we led a $15 million financing in 2025 alongside our friends at Next Frontier Capital, who introduced us to the company. The shift toward foundation model safety resulted in 5x quarter-over-quarter growth at the end of last year. The team has grown to roughly 34 people and is, in Grin’s words, “hiring like wildfire,” with notable additions including Rebekah Bastian (a founding team member at Zillow) as CMO and Dr. Alison Cerezo (an American Psychological Association AI advisory member) as Chief Science Officer.

At Foundry, we believe in the power of technology to unlock and augment human potential, especially in critical and high-stakes settings. We’ve seen this through our investments in Meru Health (improving provider capacity and patient outcomes through digitally native mental healthcare), Regard (an AI-powered physician co-pilot for diagnostics), and Ferrum Health (a governance platform for healthcare AI). mpathic extends this thread into what may be the most urgent frontier: ensuring that AI itself — now the first responder in millions of sensitive interactions — does no harm.

We’re thrilled to partner with this extraordinary team in making AI safer for the people who need it most. You can read more about mpathic and its expansion here.