What will we do with the time?
NVIDIA's Open Secure AI Alliance puts the controls around agents into focus. For funds, the useful work is a bounded workflow, limited access and a clear record of when human judgement takes over.
Hi folks,
The useful question for a fund is not just whether an agent can do the work. It is whether the workflow around it gives the team the control to trust it.
The one thing.
The controls around an agent matter as much as the model inside it
- NVIDIA and a group including Capital One, Microsoft, Palantir and Snowflake formed the Open Secure AI Alliance around identity, permissions, guardrails, logs, evaluation and isolation.
- The alliance is security-led, but the principle is useful anywhere an agent touches fund work.
What this means for a fund:
- Start with one workflow and give the agent only the access it needs.
- Name the owner, sources, review step and record. What would we trust this system to do?
In the mix.
- Kimi K3 is now an open-weight model you can actually test
- Moonshot released the full weights, making self-hosting a real option rather than a headline.
- Why it matters: more model choice does not remove the need for security, evaluation and accountability.
- HSBC is building a central AI centre, not announcing a bank-wide bot rollout
- Its Singapore centre will begin with wealth, treasury and payments, with governance and human oversight retained as work scales.
- Why it matters: centralise the guardrails, then make a few high-value workflows useful.
- The EU AI Omnibus is now in force
- It extends some timelines and expands regulatory-sandbox access.
- Why it matters: better timing does not replace ownership and records for material workflows.
Readings.
This week I came accross Can I Vibecode It? It is a playful leaderboard asking which SaaS subscriptions are one prompt away from being rebuilt. I strongly believe that as things get easier to build the most important resource is focus! I agree with the adage: if it does not make your beer taste better, don't do it.
Levercon builds custom, self-improving AI agents and systems for investment firms, with every output traced, verified and signed off.
Book a strategy call.
See where AI fits in your firm.
A short conversation about where AI could help your firm, what to build first, and what it would take.