Global Deal Struck to Have Governments Help Test AI Models Before Release

Sunak announced a deal between governments to test leading tech companies' AI models before release. Countries include the U.S., U.K., EU, Australia, Canada, etc.

The deal aims to improve AI safety by having governments, not just companies, test new models.

An international advisory panel on AI risks will be formed with 28 attending countries.

The panel will support a "State of Science" report on frontier AI risks and capabilities.

A next summit on AI safety will happen in South Korea in early 2023.