Organisations need to prove that their AI agents can shut down

Any organisation contemplating use of AI agents, especially any in national security or critical infrastructure, should give them no system access until sure that they will obey an order to stop. This is called a ...

Source: ASPI The Strategist