About
I am Alex Galle-From, a licensed Minnesota attorney working in trust and fiduciary services. This is my personal site; the projects here are personal work and are unaffiliated with my employer.
Since December 2025 I have built and released open-source tools for holding AI agents accountable — audit trails, evaluations, and consumer-rights utilities — and I wrote a model statute, the Minnesota Digital Trust & Consumer Protection Act, now at version 1.6. The work is about holding autonomous systems to account through records, liability, and verifiable duty rather than surveillance.
Fiduciary AI barely exists as a field. Banks are putting language models inside trust administration. Estate-tech companies are shipping AI-drafted documents into a liability regime written for human drafters. I work both sides of the line: by day I administer trusts inside a risk framework, and on my own time I build the evaluation, audit, and statutory infrastructure the automation will need. Everything on this site is that infrastructure.
Before law I was a musician. I moved into trust administration and, more recently, AI governance. I live and work in Minnesota.
Practice
I am a senior trust officer at a national bank, administering fiduciary relationships in the high nine-figures. Before that: estate planning and probate practice at large Minneapolis firms — wills, revocable and irrevocable trusts, gift, estate, and generation-skipping tax returns, contested probate and guardianship hearings, international estate plans with foreign co-counsel, and esoteric trust and asset-transfer work for ultra-high-net-worth clients. I hold an LL.M. in international tax law from Uppsala University. In law school I was student director of the Insurance Law Clinic.
Music
Before law I was a professional musician, a Berklee-trained violist. I played strings on Cloud Cult's "The Meaning of 8" and performed with Alison Krauss and Ron Carter.
What I’m building
Minnesota Digital Trust & Consumer Protection Act (v1.6, 2025–2026)
A model statute I drafted for AI-agent accountability. It defines bonded credentials backed by capital and strict liability for credential issuers. The framework is substrate-neutral: it regulates by accountability, not by the nature of the actor. The full text, actuarial guidance, and administrative rules are on this site.
Fiduciary Constraint Evaluation (FCE) (Apr 2026–present)
An open-source, 33-scenario evaluation of whether language models preserve fiduciary and professional constraints under pressure, with a proof-carrying runtime layer. A write-up of the pilot lives at /evals; the code is on GitHub.
AgentLedger (Dec 2025–present)
A tamper-evident audit-trail SDK for AI agents: a hash-chained log with Merkle roots and Ed25519 signatures, plus export formats for FINRA, HIPAA, and the EU AI Act. TypeScript, open source.
Contact
Email alex@alexgallefrom.io or find my code on GitHub.