Anthropic said its Claude artificial intelligence model helped identify and disrupt attempted malicious activity involving biological and conventional weapons, a Russia-linked cyber espionage campaign and Iranian propaganda networks.