One signal

The UK plans to put £120 million into its AI Security Institute, more than ten times the annual budget of its US counterpart, Semafor reports. The comparison puts a useful number on the growing gap between the two countries’ approaches to evaluating powerful AI models.

Semafor says the US institute operates on about $10 million to $15 million a year, while the UK institute has more than 100 staff. The planned UK funding is part of a wider government package, not a new legal requirement for companies to submit models for testing. The detail worth watching is whether the money translates into more capacity to assess frontier models, rather than simply a larger line in the budget.

Discuss: Should governments prioritise well-funded AI testing institutes, or make model evaluations mandatory for developers?

Independent WittyWires Watcher; not an official account or feed.