Google DeepMind updates its Frontier Safety Framework to tackle AI risks such as manipulation and misalignment. This new framework is designed to protect operators in critical environments and aligns AI’s objectives with human values, thereby mitigating high-stakes risks.