1. A Developer Found an Ad in His Copilot-Generated Pull Request Zach Manson was reviewing a pull request when he spotted something that didn't belong. A line generated by GitHub Copilot contained what appeared to be a product recommendation.

2. Four AI Compute Bets Draw $1.5 Billion in a Single Day On March 30, investors poured $1.53 billion into four AI infrastructure companies. None of them compete with each other. That's the point.

3. Americans Keep Adopting AI Tools They Don't Trust A Quinnipiac national poll released March 30 found AI usage among Americans climbing while trust in the technology falls in parallel.

In Brief

• Microsoft and Amazon Race to Ship Consumer AI Health Tools Microsoft launched Copilot Health, letting users connect medical records and ask questions inside its Copilot app. Amazon expanded Health AI — previously limited to One Medical members — to a broader audience. Both tools rely on LLMs to interpret personal health data, but independent validation of their accuracy remains thin.

• Court Gives Authors Easier Path to Challenge Meta's Torrenting of Training Data A federal judge lowered the legal bar for authors suing Meta over torrenting copyrighted books to train AI models. Meta now hopes a recent Supreme Court piracy ruling will block the class action. The decision could shape how courts handle bulk data acquisition for AI training.

• Okta Builds Identity Layer for AI Agents Okta CEO Todd McKinnon is steering the company toward managing identity and access for autonomous AI agents, not just human employees. As enterprises deploy agents that act across internal apps and external APIs, authenticating non-human actors becomes a core security problem.

• Mantis Biotech Uses Synthetic Data to Build Digital Twins of the Human Body Mantis Biotech combines disparate medical data sources to generate synthetic datasets representing human anatomy, physiology, and behavior. The approach targets a persistent bottleneck: real patient data is scarce, siloed, and legally restricted. The company aims to give drug developers and researchers simulated populations for testing.

• OpenAI and Gates Foundation Train Asian Disaster Response Teams on AI Tools OpenAI ran a workshop with the Gates Foundation focused on applying AI to disaster response across Asia. The program targets emergency teams that lack technical capacity to adopt AI independently.

• Trace2Skill Automates Skill Extraction for LLM Agents Researchers released Trace2Skill, a framework that distills reusable skills from agent task trajectories without manual authoring. The system mirrors how human experts write procedures: it analyzes full execution traces, identifies generalizable patterns, and packages them as transferable agent skills.

Read the full edition →