Igor Ivanov comments on On closed-door AI safety research

[Igor Ivanov]

This was written in the Claude 4 system card. Made sense to test Claude and not other LLMs