Seattle startup uses clinical expertise to make AI models safer and reduce dangerous responses

As millions of users — including large numbers of young people — increasingly turn to AI chatbots as their first-line “counselors” and confidants, Seattle-based startup mpathic is stepping in to ensure those digital agents don’t provide dangerous advice when it matters most.

The company, founded in 2020 in a bid to bring more empathy to corporate communication, announced Monday that it is expanding to support foundational model developers and LLM-powered application teams.

The goal is to bring mpathic’s software to a broader set of AI developers and enterprise partners as AI becomes more of an interface for mental health and medical support.

“We are essentially producing eval sets or training data sets to make models more safe for vulnerable users, like kids or people with mental health problems, people in crisis,” said mpathic co-founder and CEO Grin Lord, a board-certified psychologist and NLP researcher.

The startup is drawing on its years of work in clinical trials and hospital settings, helping AI teams stress-test model behavior before deployment, evaluate responses, and monitor live interactions with safeguards that can flag, redirect, or intervene when needed.

“It’s kind of similar to people that create synthetic data for visual AI,” Lord said. “It’s not every day that a child is going to run in front of a Waymo, but we can simulate that 10,000 ways with synthetic data. That’s basically what we’re doing, but from a psychological angle with language.”

In one early engagement, mpathic said its clinician-led program helped a model builder slash undesired or dangerous responses by more than 70%.

To fuel its expansion, mpathic raised an additional $15 million in 2025, led by Foundry VC. The company says the move toward foundational safety resulted in 5X quarter-over-quarter growth at the end of last year.

While Mpathic got its start building software to analyze conversations happening in corporate texts, emails, audio calls, and more, it has been developing models for high-risk clinical situations since 2021. Today, the scale of the startup’s “human-in-the-loop” infrastructure includes a global network of thousands of licensed clinical experts. It is onboarding hundreds more weekly to keep pace with demand.

“It’s a lot different company than it was even a few quarters ago,” Lord said.

Lord, a finalist for Startup CEO of the Year at the 2023 GeekWire Awards, calls herself a “techno optimist” and “realist” when it comes to AI, adding that she possesses a “radical acceptance” of the technology’s usefulness.

“It doesn’t surprise me at all that if there’s something that’s available 24/7, that acts like a therapist, you’re going to talk to it and use it. And that could be better than nothing,” she said. “I think the potential for this technology to have really positive impact is super high. I think we can train both humans and AI to listen accurately and well and not create harm.”

Without naming specific companies or models, Mpathic confirmed it is working with leading foundational AI model developers serving tens of millions of users. The startup also has clinical partners including Panasonic WELL, Seattle Children’s Hospital, Transcend and others.

Mpathic, which employs roughly 34 people and is “hiring like wildfire,” according to Lord, has also grown its leadership team with the addition of chief marketing officer Rebekah Bastian (Zillow, AllTrail, Glowforge); and chief science officer Alison Cerezo (American Psychological Association AI advisory member).