1 article tagged “confidential-computing”, most recent first.
Google DeepMind is testing a cryptographic way to keep both model weights and evaluation prompts hidden during independent AI assessments.