Reports emerged Wednesday that the White House is preparing an executive order directing federal agencies to challenge or block state-level AI regulations.
Seattle entrepreneur Joe Braidwood sees the news as a major opportunity.
Braidwood is CEO and co-founder of Glacis, a new startup backed by the AI2 Incubator that is building software to help companies prove that their AI safety measures are executed as intended. Glacis creates tamper-proof “receipts” for every AI decision, allowing companies to prove their safety systems actually ran.
“Think of it as a flight recorder for enterprise AI,” Braidwood said.
Braidwood said the potential White House order to block state AI laws transforms Glacis from a startup just getting off the ground into “infrastructure necessity.” In an environment where the Justice Department would sue states that pass AI rules, a neutral, platform-agnostic trust layer could become increasingly relevant.
Glacis’ origins are rooted in regulatory complexity.
Braidwood, a longtime tech marketing leader, recently shuttered Yara, his year-old startup that aimed to use AI to improve mental wellness. He cited Illinois regulations that made AI therapy “effectively uninsurable.”
In a LinkedIn post that has since gone viral, Braidwood explained the decision to close Yara and open-source a set of safety prompts he had developed.
He wrote that Yara closed after he realized AI became “dangerous” when interacting with people facing deep trauma or suicidal ideation — not just inadequate. The experience, he said, showed “where the boundaries need to be,” and demonstrated how startups working in high-risk AI categories face unmanageable liability and regulatory pressure.
After the post, regulators, clinicians, engineers, founders and insurance executives reached out — many pointing the same problem: when AI systems make decisions, no one can independently verify whether safety policies actually fired.
That clarity became the seed for Glacis.
Every time an AI model answers a question or takes an action, Glacis creates a signed record showing the input, the safety checks that ran, and the final decision. The record can’t be altered and takes less than 50 milliseconds to generate. Regulators and insurers can verify these receipts without seeing any personal data, and Braidwood said insurers believe this could finally make it possible to insure AI systems that can prove they followed the rules.
Braidwood co-founded Glacis with Dr. Jennifer Shannon, a psychiatrist and adjunct professor at the University of Washington.
The company is currently in private beta with digital health customers, including nVoq, and is targeting healthcare, fintech, and insurance sectors. It’s also part of Cloudflare’s Launchpad program.
Braidwood was previously chief strategy officer of Vektor Medical. He also co-founded social TV platform Scener and was chief marketing officer at SwiftKey.
Shannon has been a psychiatrist for nearly two decades. She was also a medical director at Cognoa and serves on the AI Resource Committee for the American Academy of Child and Adolescent Psychiatry.
