The brief · 24 April 2026
Three angles, three scales: the gap in pediatric medical data, a million open-source tokens from DeepSeek, and the DOJ stepping up for xAI in Colorado.
An AI trained mostly on adults performs worse on pediatric scans, and nobody catches this until it's actually been tested. If your pediatrician or hospital is already using an AI-assisted imaging tool, it's worth asking what data it was trained on.
practitioners › 203 public medical imaging datasets analyzed: 33% had no age metadata, and children made up under 2% of patients when age was recorded (Nature Health, April 24, 2026).
www.nature.com →An open-source model with context this long changes what you can feed it in one pass: a full report, an entire codebase, several books at once. No more chunking your documents before sending them in.
practitioners › Open weights on Hugging Face, API available, V4-Pro at 1.6T parameters and V4-Flash at 284B, preview versions live since April 24, 2026.
DeepSeek →xAI was already suing Colorado over its AI law (SB24-205). Now the DOJ has officially joined the case, and that changes the stakes entirely: this is no longer a company versus a state, it's the federal government versus state law. If this succeeds, anti-discrimination requirements for AI models across the US take a serious hit.
practitioners › DOJ is invoking the Equal Protection Clause and specifically targeting the law's DEI exemptions; a joint motion for an emergency injunction against SB24-205 enforcement was filed on April 24, 2026.
www.justice.gov →