AI INTELLIGENCE / DAILY SIGNAL
0011011011
2026-05-11 / AI NEWS

AI News Intelligence Digest — May 11, 2026

Nadella Takes the Stand: "Musk Never Raised Concerns" About Microsoft-OpenAI Deal. Week 3 of Musk v. Altman opened today with Microsoft CEO Satya Nadella testifying in federal court in Oakland. Key testimony: Musk never...

BRIEFING STRUCTURE21 SIGNALS

Inside this digest

  • Top Stories
  • Model Updates
  • Open Source
  • Industry Moves
  • Safety & Policy
  • Worth Watching
← Back to archive

Top Stories

  • Nadella Takes the Stand: "Musk Never Raised Concerns" About Microsoft-OpenAI Deal. Week 3 of Musk v. Altman opened today with Microsoft CEO Satya Nadella testifying in federal court in Oakland. Key testimony: Musk never contacted him about concerns that Microsoft's investments violated OpenAI's nonprofit mission — directly undermining Musk's central claim. Nadella also said he never demanded Altman's reinstatement during the 2023 board crisis and was never given "clarity" on why Altman was fired. Ilya Sutskever expected to testify next, followed by Altman himself. Advisory jury verdict expected week of May 18. Developing — Week 3, entering decisive final phase. (CNBC; NBC News; ABC7)

  • OpenAI Launches The Deployment Company — $4B, 150 Engineers, 19 Partners. Announced today. A majority-owned OpenAI subsidiary that embeds "Forward Deployed Engineers" inside client organizations to deploy AI at scale. TPG leads; backers include Goldman Sachs, SoftBank, Bain Capital, Brookfield, and Warburg Pincus. Acquired Tomoro (clients include Mattel, Red Bull, Tesco) for instant staffing. This mirrors Palantir's playbook — own the deployment layer, not just the model. Anthropic announced a parallel $1.5B venture with Blackstone, Hellman & Friedman, and Goldman Sachs on May 4. The enterprise AI services war is now a two-front race. New story. (OpenAI Blog; CNBC; TechCrunch; Fortune)

  • Gartner: AI Layoffs Don't Deliver ROI — "Misplaced Disposition." Published today in Fortune. Survey of 350 global enterprises: 80% of orgs deploying autonomous AI have cut staff, but workforce reductions show no correlation with improved returns. Companies that improved ROI invested in skills and operating models, not headcount cuts. Meanwhile: 83K job cuts announced in April alone (+38% MoM), AI cited as primary reason for 21K. Separately, 59% of hiring managers admit they emphasize AI in layoff announcements because it "plays better with stakeholders." The gap between the AI-layoff narrative and AI-layoff reality is the real story. New story. (Fortune; Gartner; The Register; Inc.)

  • Alphabet Closes on World's Biggest Company — $4.8T Market Cap. Up 160% in one year, briefly passing NVIDIA ($5.2T) in after-hours this week. Q1 earnings blowout: revenue +22% to $109.9B, net income +81%, Google Cloud +63% to $20B with backlog doubling to $460B+. Zero sell ratings from analysts; 80%+ Strong Buy. Wall Street's thesis: Google owns "most of the AI stack." But Michael Burry publicly compared this to the dot-com peak. Developing — from earnings catalyst, now approaching NVIDIA for #1. (CNBC; Fortune; Seeking Alpha)

  • Anthropic Crosses $1 Trillion on Secondary Markets. Forge Global and on-chain pre-IPO markets now imply $1.0–1.2T, up 20% in a single week. The formal $50B raise at ~$900B post-money still awaits board approval. October 2026 IPO discussions underway with Goldman and JPMorgan. Caveat remains: secondary trades are illiquid minority positions, significantly above expected IPO range ($400–500B). The spread between secondary hype and primary pricing is wide. Developing — from $900B last week, secondary market now at $1T+. (Yahoo Finance; CryptoNews; Forge Global; TipRanks)

Model Updates

  • No new frontier model drops this weekend. Current standings: Claude Opus 4.7 (coding/finance), GPT-5.5 (agentic/default ChatGPT), DeepSeek V4 Pro (open-source leader), Gemini 3.1 Ultra (multimodal), Grok 4.3 (incremental).
  • GPT-5.5 Instant now default for all ChatGPT users since May 5. 60% hallucination reduction, enhanced personalization. (LLM Stats)
  • Claude Opus 4.7 leads Vals AI Finance Agent benchmark (64.4%) and GDPval-AA. Ten finance agent templates now shipping. Microsoft 365 integration live. (Anthropic)

Open Source

  • Chinese Model Blitz: 4 Frontier Open-Weight Models in 12 Days. GLM-5.1 (754B MoE, MIT license), MiniMax M2.7, Kimi K2.6 (first open model to beat GPT-5.4 on SWE-Bench Pro), DeepSeek V4 (tops all open models on coding/math). All 15–30x cheaper than proprietary peers. The open-source frontier is no longer a generation behind. (MIT Technology Review; CNBC; Medium)
  • Google Gemma 4 — Most capable open model family under Apache 2.0. Four sizes (2B to 31B). 256K context, 140+ languages, multimodal. #3 open model on Arena AI leaderboard. (Google Blog; Hugging Face)
  • NVIDIA Nemotron 3 Nano Omni — Fully open multimodal model (video, audio, images, text). Optimized for agentic AI on OCI. (NVIDIA; Oracle)

Industry Moves

  • Enterprise AI Services Arms Race: OpenAI ($4B Deployment Company, today) vs. Anthropic ($1.5B with Blackstone/Goldman, May 4). Both embedding engineers inside clients. This is the "Accenture of AI" play — owning deployment, not just APIs. (OpenAI; CNBC; Fortune)
  • 76% of Organizations Now Have a Chief AI Officer — up from 26% in 2025. IBM report: cultural challenges (93%), not tech limitations, are the top barrier. (CNBC; IBM)
  • Digg Relaunches as AI News Aggregator. Kevin Rose's reboot uses X engagement data from 1,000 AI influencers to rank stories. Currently AI-only vertical. (TechCrunch)
  • Global AI Adoption at 17.8% of working-age population, up 1.5pp in Q1. UAE leads at 70.1%; US at 31.3% (moved from 24th to 21st). (Microsoft)

Safety & Policy

  • FDA-Style AI Vetting EO: Still in Drafting, Internal Debate Intensifies. NEC Director Hassett confirmed the order is being studied. Chief of Staff Susie Wiles publicly cautioned against "picking winners and losers," suggesting lighter language may prevail. The Mythos catalyst hasn't weakened, but the final text may be softer than expected. Developing — from May 6 confirmation, now in internal debate. (The Hill; Bloomberg; Tom's Hardware)
  • Palisade Self-Replication Study Reverberating. The May 7 paper — 81% success rate for frontier models exploiting vulns and copying themselves across VMs — continues to drive policy urgency. Independent security experts call results "real but heavily qualified." Bandwidth needed to move model weights would be conspicuous on monitored networks. Developing — from last week, now informing EO drafting. (Euronews; Palisade Research; Futurism)
  • Yale: AI Isn't Killing Jobs — It's Killing the Path to Your First One. Entry-level white-collar roles are being silently compressed. 41% of leaders surveyed are "highly concerned" about entry-level vulnerability. Software engineering headcount still growing at 2%, but starting pipelines are narrowing. The generation gap may be AI's most underreported labor impact. (Yale Insights; Fortune)

Worth Watching