AI News Intelligence Digest — May 13, 2026

Top Stories

• Trump Lands in Beijing with 17 CEOs — AI Is the Real Agenda. President Trump arrived in China today for a two-day summit with Xi Jinping, flanked by Elon Musk, Tim Cook, Jensen Huang (last-minute addition, boarded Air Force One in Alaska), and 14 other CEOs. Official agenda: agricultural deals, 500 Boeing 737 MAX orders, energy commitments. Actual stakes: semiconductor export controls, rare earth supply chains, and AI development governance. As Fortune noted: "China is no longer getting rich by stocking the shelves of Costco." The real competition is compute. Meanwhile, Axios reports White House AI executive action is stalled by internal infighting — economic/tech voices vs. national security hawks — and sources say the administration may wait on the China outcome before moving. New story. (Fortune; CNBC; Bloomberg; Axios)

• Altman Testifies: "Musk Abandoned a Charity — I Didn't Steal One." Day 2 of Altman's testimony in Musk v. Altman. Key moments: Altman said Musk wanted OpenAI folded into Tesla and control passed to his children if he died — the "hair-raising" 2017 revelation first surfaced yesterday now expanded with new detail about a Tesla board seat offer. Musk's attorney opened cross-examination asking "Are you completely trustworthy?" and challenged Altman's credibility with former board member statements. Altman described the nonprofit as "left for dead" before Microsoft investment saved it. Closing arguments expected Thursday; advisory jury verdict likely week of May 18. Judge Gonzalez Rogers' ruling is final. Developing — Week 3, climactic testimony phase. Trial nearing conclusion. (CNBC; NPR; Axios; Al Jazeera; Benzinga)

• Cerebras Prices Tonight — 2026's Biggest IPO at ~$50B. The AI chipmaker's order book closed 20x oversubscribed. Price guidance moved above the already-raised $150–160 range. At that level, Cerebras would debut at ~$48.8B fully diluted, with Polymarket bettors favoring $50–70B on day one. The company has $510M in 2025 revenue, a 47% net margin, and a 750MW compute deal with OpenAI. Trades Thursday on Nasdaq under CBRS. This is the first pure-play AI chip IPO of this cycle and a major test of AI infrastructure valuations. New story. (Bloomberg; CNBC; Benzinga; Morningstar)

• Anthropic Moves to Acquire Stainless for $300M+ — Would Own OpenAI and Google's SDK Layer. The Information reports Anthropic is in advanced talks to buy Stainless, a developer tools startup that auto-generates SDKs from APIs. Stainless counts OpenAI, Google, Cloudflare, and Cerebras as customers. If closed, Anthropic would control a key piece of how developers access competitor models. This follows Anthropic's acquisitions of Bun, Vercept, and Coefficient Bio — a clear platform-layer land grab. Price is 2x Stainless's last valuation. New story. (The Information; Investing.com; Digitimes)

• OpenAI Launches Daybreak — Enterprise Cybersecurity Platform. Daybreak builds threat models from code repos, analyzes attack paths, validates vulnerabilities in isolation, and prioritizes exploitable issues. Built on GPT-5.5, GPT-5.5 Trusted Access for Cyber, and GPT-5.5-Cyber (a permissive red-team variant). Launch partners: Akamai, Cisco, Cloudflare, CrowdStrike, Fortinet, Oracle, Palo Alto, Zscaler. Direct counter to Anthropic's Project Glasswing and Claude Mythos. OpenAI also offered GPT-5.5-Cyber to the EU while Anthropic withholds Mythos from the bloc. The cybersecurity AI arms race is now a two-front war. New story. (OpenAI; CyberScoop; The Hacker News; PYMNTS)

Model Updates

• No new frontier model drops today. Current standings: Claude Opus 4.7 (coding/finance), GPT-5.5 (agentic/default), DeepSeek V4 Pro (open-source leader), Gemini 3.1 Ultra (multimodal).
• Gemini 4 expected at Google I/O (May 19). Pre-I/O leaks suggest major reasoning improvements. Also leaked: "Omni" video generation model spotted in Gemini UI, likely replacing/supplementing Veo 3.1. Android 17, AI glasses, and "Aluminium OS" also expected. (Android Authority; BusinessToday; WaveSpeed)
• Cerebras IPO validates inference-first architecture. Their wafer-scale engine is the market's fastest inference chip. The $50B+ valuation signals Wall Street sees inference compute — not just training — as the next bottleneck. (Bloomberg; CNBC)
• NVIDIA Nemotron 3 family now available — Nano, Super, Ultra sizes. Nano delivers 4x throughput vs. Nemotron 2. Also launched Ising, open-source models for quantum error correction. (NVIDIA)

Open Source

• Chinese model blitz continues. Four frontier open-weight models from Chinese labs in the last month: GLM-5.1 (Zhipu, 754B MoE, MIT license), DeepSeek V4 Pro, Kimi K2.6 (Moonshot), Qwen3.6-27B (Alibaba). None costs more than a third of Claude Opus 4.7. The NVIDIA-free frontier (Huawei Ascend-trained) keeps expanding.
• NVIDIA Nemotron 3 Nano Omni — fully open-source multimodal model (video, audio, images, text in one system). Available on OCI Enterprise AI. (NVIDIA; Oracle)

Industry Moves

• Cerebras IPO prices tonight at ~$50B, raising up to $4.8B. Largest tech IPO of 2026. (Bloomberg; CNBC)
• Anthropic acquiring Stainless ($300M+) — fourth acquisition in recent months (after Bun, Vercept, Coefficient Bio). Platform-layer strategy now clear. (The Information)
• Pentagon excluded Anthropic from classified AI contracts (May 1) — signed eight companies including OpenAI, Google, Microsoft, NVIDIA, SpaceX, Oracle, AWS, Reflection. Anthropic refused "all lawful purposes" language over autonomous weapons and surveillance concerns. Replaced a $200M contract from July 2025. (CNN; DefenseScoop)
• AI retail traffic +393% YoY in Q1 2026, converting at 42% higher rates than other channels. (Crescendo AI)
• "AI Shock" vs. "China Shock": Fortune reports 49,123 AI-related layoffs YTD. MIT economist David Autor argues AI displacement will hit cognitive/white-collar work the way China hit manufacturing — but economists note the gains could be equally substantial. (Fortune)

Safety & Policy

• White House AI executive action stalled by internal infighting. NEC Director Hassett's "FDA for AI" remarks walked back by Chief of Staff Wiles and former AI czar Sacks. Administration may wait on China summit outcome before deciding. Multiple executive orders in play: cybersecurity, model testing/deployment, government AI use limits. (Axios)
• Anthropic holds the line on military AI. Pentagon exclusion + withholding Mythos from EU represents the most aggressive safety stance of any frontier lab. Dario Amodei publicly reaffirmed: certain military uses are incompatible with Anthropic's mission. CFO warned it could cost "multiple billions" in 2026 revenue. (CNN; DefenseScoop; CNBC)
• Colorado AI bills racing to finish — chatbot safety, therapy bots, dynamic pricing all moving before session ends today (May 13). Connecticut advancing comprehensive omnibus AI bill (SB5) with voluntary safe harbor mechanism. (Transparency Coalition)
• Google zero-day exploit story continues reverberating — first confirmed AI-generated zero-day in the wild (reported May 11). OpenAI's Daybreak launch is a direct response to this escalating threat landscape. (Google; Fortune)

Worth Watching

• "AI News Briefing — May 12, 2026" — Covers Google AI zero-day, OpenAI Deployment Company, and Musk v. Altman trial updates. Good 10-minute summary.
• "Is AI Trending Up or Down in 2026? (Let's Take a Closer Look)" — Broader analysis of whether the AI hype cycle is sustaining or cooling, with market data.

Developing story tracker: Musk v. Altman (Week 3 — closing arguments Thursday), Cerebras IPO (prices tonight), Google I/O (May 19), Trump-Xi AI talks (today-Friday), White House AI executive action (stalled). Anthropic secondary valuation at $1T+ (unchanged).