Announcement
Japan AISI publishes cyber-capability test results for Claude Opus 4.8 and GLM-5.2
Japan AISI says the tests are a first step toward creating its own benchmarks for evaluating AI safety.
TLDR
Japan AISI says it tested the cyber capabilities of Claude Opus 4.8 and GLM-5.2 and published the results. It aims to create safety benchmarks and evaluate AI safety itself. The organization also announced the launch of its X account.
Combined views
—
1 Source, first seen ago
