In a recent report, Anthropic said terror groups in the middle east and Africa have tried to use its "Claude" chatbot to develop weapons and attack plans.