Technical Governance·Exploring·Last reviewed May 1, 2026
This page is a stub. I’ve marked the territory but haven’t written my views here yet. The headings below are placeholders — the actual beliefs, uncertainties, and evidence are still in my notes. If you want my current take on this topic before it lands here, get in touch.
Where I currently stand
<Headline view: structured access is a real governance instrument but the field has been imprecise about what counts as "access" — query access, fine-tune access, weight access, training-data access, and training-process access produce wildly different evidence. The most useful work specifies what evidence each access tier actually generates.>
Current beliefs
- Query-only access is sufficient for capability evals but not for safety-property evals. ~XX% — <why>.
- Audit standards have to specify the access tier before they specify the test. ~XX% — <why>.
- <Claim about indemnification / safe-harbour for auditors.> ~XX% — <why>.
Uncertainties
- What is the smallest access bundle that produces a regulator-defensible audit? Why it matters: defines the negotiating floor for any structured-access regime.
- Can auditors retain useful independence inside a regime where labs have meaningful veto over access terms? Why it matters: determines whether third-party audit is a real check or a compliance theatre.
What would update me
- A high-quality public report from a structured-access audit demonstrating clear, novel safety findings.
- A high-profile audit producing only platitudes despite high-tier access would suggest the regime needs redesign.
Recent reading
- <date> — <title> — <takeaway>.
Related writing
No essays tagged with this topic yet.