Earlier quoted context omitted.
>We should give up with the idea of databases which are 'open' to the public, but you have to pay to access, reproduction isn't allowed, records cost pounds per page, and bulk scraping is denied. That isn't open. How about rate limited?
No. Open is open. Beyond DDoS protections, there should be no limits. If load on the server is a concern, make the whole database available as a torrent. People who run scrapers tend to prefer that anyway. This isn't someone's hobby project run from a $5 VPS - they can afford to serve 10k qps of readonly data if needed, and it would cost far less than the salary of 1 staff member.
I could have my mind changed if the public policy is that any public data ingested into an AI system makes that AI system permanently free to use at any degree of load. If a company thinks that they should be able to put any load they want on public services for free, they should be willing to provide public services at any load for free.