Verification for regulated AIOur community · C Combinator

Collaboration

Build with the people
who carry the risk.

Legis Labs welcomes collaboration with financial institutions, insurers, regulators, researchers and technology providers around practical, testable agent-authority problems.

Banks & payment providers

Define a permitted workflow, an action boundary and the evidence required by your risk and engineering teams.

Proposed collaboration: a scoped design-partner pilot

Insurers & brokers

Explore what authority records, control tests and incident evidence could support an underwriting decision.

Proposed collaboration: an evidence workshop

Regulators & public bodies

Examine transparent control methods and failure cases using synthetic or appropriately governed data.

Proposed collaboration: technical dialogue and eligible testing programmes

Academics & standards groups

Investigate delegation, policy enforcement, evaluation methods and interoperability across implementations.

Proposed collaboration: reproducible research and review

These are invitations to collaborate. No regulatory endorsement, institutional partnership or insurance acceptance is implied.

Our community initiative

C Combinator

A place for thoughtful discussion and practical work across finance, law, healthcare and AI. The initiative brings people together through gatherings in Cambridge and London.

Explore current events, participation and community opportunities on the C Combinator website.

Visit C Combinator ↗
Open questions.Shared learning.Practical work.

Proposed working programme

Give each conversation
a useful output.

A mandate and evidence working group

Compare what different institutions need to know before an agent can act.

A control-testing workshop

Use synthetic cases to explore escalation, revocation and failed assumptions.

A published methods note

Document the method, findings and limitations with contributors’ permission.

Work with Legis Labs

Bring a question worth testing.

Tell us your role, the workflow and the decision you want better evidence to support.

Discuss an evaluation