New benchmark finds top Artificial Intelligence language models fall short in mental health care
Sword Health’s MindEval simulates multi-turn patient-clinician conversations and benchmarks 12 frontier models including GPT-5, Claude 4.5, and Gemini 2.5, finding average clinical performance stayed below 4 on a 1-6 scale.
Rising Artificial Intelligence prompt injection attacks pose new security risks, warns NCSC
The National Cyber Security Centre warns that rising Artificial Intelligence prompt injection attacks let attackers embed hidden commands in user input, forcing organizations to update defenses, monitoring and testing.
White House launches Genesis Mission to accelerate Artificial Intelligence-enabled scientific discovery
The White House issued an executive order on November 24, 2025 establishing the Genesis Mission, a Department of Energy-led initiative to build an American Science and Security Platform to accelerate Artificial Intelligence-enabled scientific discovery and automated manufacturing. The order sets binding near-term deadlines for DOE to identify priority challenges, inventory computing resources, and integrate datasets and model assets.
NVIDIA gets H200 export license for China, 25% of revenue to the U.S. government
The Trump administration approved a framework allowing NVIDIA to resume H200 exports to China under Commerce Department licensing, with 25% of sales proceeds directed to the U.S. government. The H200 remains a powerful option for large-scale Artificial Intelligence workloads, balancing market access with regulatory controls.