AI News Briefing — February 10, 2026

The European Commission has threatened action against Meta for blocking rival AI chatbots from its WhatsApp Business platform, arguing it constitutes an abuse of dominant market position under EU antitrust rules. This signals growing regulatory scrutiny of AI distribution chokepoints.

Goldman Sachs is partnering with Anthropic to deploy autonomous AI agents powered by Claude for complex back-office operations including accounting, compliance, and client onboarding. The bank's CIO says the technology has exceeded expectations in handling tasks previously deemed too complex for automation.

With the last major US-Russia nuclear arms treaty having expired, experts are debating whether satellite surveillance combined with AI monitoring could serve as a substitute for traditional nuclear treaties. The proposal remains controversial among arms control specialists.

Researchers from Harvard and Stanford have released Ordered Action Tokenization (OAT), a framework that enables LLM-style autoregressive scaling for robotics by solving the long-standing challenge of converting continuous robot actions into discrete tokens. The approach could unlock GPT-style scaling laws for robotic control.

Nearly a year after New York state began requiring companies to disclose when technological innovation or automation causes job losses, no company has made such a disclosure. The gap raises questions about enforcement and the true pace of AI-driven workforce displacement.

Microsoft Research introduces OrbitalBrain, a framework that turns nanosatellite constellations into distributed ML training systems, enabling model training directly in space rather than waiting for slow downlinks to Earth. The system uses inter-satellite links and predictive scheduling of power and bandwidth.

An overview of the AI legal landscape in 2026 notes that while more settlements may come, fundamental questions about fair use and copyright infringement in AI training remain unresolved. The legal uncertainty continues to shape how AI companies approach data and model development.

A startup has introduced a 'Large Tabular Model' specifically designed to handle structured spreadsheet data, addressing an area where traditional large language models consistently underperform. The tool targets enterprise data analysis workflows.