AI safety scores, infrastructure bets and legal pressure reshape the week

AI’s largest labs faced new scrutiny after the Future of Life Institute’s 2026 AI Safety Index gave Anthropic the top grade at C+, with OpenAI and Google DeepMind at C, Meta at D+, and xAI, DeepSeek, and Mistral effectively failing. The index assessed risk management, transparency, governance, and whether labs honored safety commitments, warning that several major companies had softened earlier promises.

Governments and investors continued to pour money into the physical layer of AI. South Korea announced a decade-long plan worth 1,350 trillion won, roughly $880 billion, including $518 billion for memory chip manufacturing and about $550 billion for AI data centers. Meta also expanded its enterprise push with a Business Agent Platform and Meta Compute, while Blackstone, Apollo, and KKR backed a $5.34 billion power deal for data centers.

Legal, labor, and geopolitical pressures also intensified. The New York Times and other publishers sought sanctions against OpenAI over allegedly withheld training-data evidence, Mews cut roughly 15 percent of its workforce and cited AI efficiency, and the EU moved toward pre-market model testing with ENISA. Attention was also turning to July 17, when Gemini 3.5 Pro was expected to coincide with Shanghai’s 2026 World Artificial Intelligence Conference.