Over time, there might be an increasingly large gap between insider model access and outsider model access. By insiders, we mean employees at the frontier lab. By “outsiders”, we mean external safety researchers, third-party auditors, and other actors trying to make the future go well.
We call this a model access gap — and when the gap is small, we call this model access parity.
Questions:
How important is model access parity for things going well?
How likely is model access parity by default?
What can we do to ensure model access parity?
What will be the main bottlenecks on model access parity? E.g. lab incentives, outsider funding, government regulation, etc.
How does model access parity depends on timelines, takeoff speeds, capability profiles, etc?
Which orgs are highest priority for model access?
What should governments do, to ensure model access parity?
What should safety-minded lab employees do, to ensure model access parity?
What should outsiders do if they don’t achieve model access parity? E.g. join a lab, switch to low-uplift activities, etc.
What would be the downsides of model access parity, and how can we mitigate them?