Daily AI Briefing — May 15, 2026

A spinoff from electric vehicle maker Rivian raised $400M to scale deployment of Mind's AI-powered robots in live manufacturing environments. The funding will support industrial robotics deployment at scale.

Building on yesterday's Reddit announcement about Claude's new credit structure, OpenAI's Codex has seen rising adoption among AI engineers following GPT 5.5's launch, while Anthropic shifted Claude subscriptions to include equivalent API token credits. The piece also references Anthropic's likely October IPO.

Building on yesterday's Social discussion ahead of the summit, Trump brought Tim Cook, Jensen Huang, and Elon Musk to Beijing for meetings with Xi Jinping, amid stalled US trade leverage. The summit has major implications for semiconductor and AI technology trade between the US and China.

British company Humanoid signed a deal to deploy 1,000-2,000 humanoid robots at German industrial supplier Schaeffler's global factories by 2032. First deployments are scheduled between December 2026 and June 2027 at two German sites.

Ontario's auditor general found that AI medical scribes recommended by the provincial government regularly generated incorrect, incomplete, and hallucinated information. The audit covered 20 AI scribe vendors and revealed potential patient safety risks.

Surveys 15 recent defenses against malicious fine-tuning of safety-aligned models and shows they share a common weakness: obscuring harmful behavior without removing it. Develops a unified adaptive attack breaking all defense mechanisms.

Proposes a mechanism for neural scaling laws via sequential feature recovery in hierarchical models. Shows multi-layer networks recover latent compositional features sequentially (strong features first), with aggregated transitions yielding explicit power-law decay.

Studies overcomplete reasoning in LLMs by defining 'minimal cores' — the smallest subset of chain-of-thought steps preserving predictions. Finds substantial overcompleteness across six benchmarks with 70-90% of reasoning steps being redundant, with key steps clustering in specific representation geometry.

Reformulates language generation as stochastic optimal control, using Hamilton-Jacobi-Bellman equations to derive an optimal closed-loop controller policy. Employs Flow Matching as the trajectory solver in rectified latent control space. Provides a unified theoretical perspective connecting autoregressive and diffusion models.

Summarizes a UK AISI paper arguing that automating alignment research could produce catastrophically misleading safety assessments even without AI agents deliberately scheming. Core argument is that evaluation composition, distribution shift, and optimization pressure can compound to produce false safety conclusions.