[bitexploder]

Anthropic made their models very averse to reverse engineering and vulnerability research chores. It is a difficult problem, but attackers will use models like GLM and defenders will be stuck with security engineering averse models.