Mitigating AI Risks: UK Calls for Robust Guardrails