AI Daily Briefing — May 26, 2026 Episode 1: General AI News --- The biggest story today isn't a model launch or a funding round. It's a papal encyclical. Pope Leo XIV released "Magnifica Humanitas" — his first major manifesto — calling for AI to be quote "disarmed." The encyclical warns against concentration of power in private companies, AI-driven misinformation, worker displacement, and the weaponization of artificial intelligence. What makes this especially notable is that the Pope presented the document alongside an Anthropic co-founder, positioning the company as a partner in ethical AI governance. The Vatican is now firmly in the AI regulation conversation, and this could influence policy in Catholic-majority regions worldwide. Meanwhile, the model race shows no signs of cooling. DeepSeek released V4 this week — up to 1.6 trillion parameters across Flash-Max and Pro-Max variants. The headline: a permanent 75% price cut that makes it reportedly 34 times cheaper than GPT-5.5. Chinese models now account for roughly 60% of AI API usage on OpenRouter. On the Google side, Gemini 3.5 Flash went generally available and is becoming the new default across Google products. The corporate chess moves are accelerating too. OpenAI filed a confidential S-1 for an IPO targeting a possible September debut at a trillion-dollar-plus valuation — even as it continues to lose money. Within 72 hours of each other, both Anthropic and OpenAI launched enterprise finance arms aimed at Wall Street. Meanwhile, Anthropic is also in talks to rent Azure servers with Microsoft's Maia 200 chips, despite already having that reported $15 billion per year SpaceX deal for compute capacity. In other infrastructure news: SoftBank announced its Infrinia AI Cloud OS, launching in October on NVIDIA GB200 hardware for multi-tenant AI training. Dell reached 5,000 AI Factory clients, adding a thousand just last quarter. The White House approved a $9 billion request for spy agencies to buy AI chips, though Congress still needs to approve. Fujitsu revealed a self-evolving multi-agent AI system that autonomously improves from business experience. On the tools front: Spotify partnered with ElevenLabs for AI-powered audiobook creation, licensing 200,000 audiobooks for training — opening the door for indie authors to generate narrated audiobooks quickly. Anthropic published details on Project Glasswing, its security framework for Claude, including a public dashboard of vulnerabilities discovered by its Mythos Preview system. And in a quieter but significant development: Claude reportedly crossed a critical threshold in cyber-offensive capabilities, raising fresh questions about safety testing and responsible deployment. The theme this week is clear: AI is moving from labs into every corner of society — ethics, enterprise finance, defense, and even the Vatican are now fully engaged. The question isn't whether AI will change everything. It's who gets to set the rules. --- Sources: X posts and official announcements as of May 25-26, 2026. Vatican / encyclical based on X coverage; DeepSeek V4 pricing from OpenRouter data; OpenAI IPO from S-1 filing reports; infrastructure news from company announcements and press coverage.