AI INTELLIGENCE / DAILY SIGNAL
0011011011
2026-07-13 / AI NEWS

AI News Intelligence Digest — July 13, 2026

Musk and Altman turn Apple's lawsuit into a public brawl. Two days after Apple sued OpenAI alleging systematic trade secret theft (400+ ex-Apple employees, hardware chief Tang Tan directing staff to bring "actual parts"...

BRIEFING STRUCTURE21 SIGNALS

Inside this digest

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

Top Stories

  • Musk and Altman turn Apple's lawsuit into a public brawl. Two days after Apple sued OpenAI alleging systematic trade secret theft (400+ ex-Apple employees, hardware chief Tang Tan directing staff to bring "actual parts" to interviews), Musk and Altman went at each other on X. Musk revived the "Scam Altman" nickname; Altman fired back: "homeboy you're the one selling public market investors on short-term space datacenters." The timing is brutal for OpenAI — IPO prep amid active litigation. Musk's own jury loss in the original OpenAI lawsuit makes his pile-on ironic. (CNBC — "Elon Musk and Sam Altman spar on X after Apple files OpenAI lawsuit"; Fortune; Quartz)

  • GPT-Live ships — voice AI goes full-duplex. OpenAI launched GPT-Live, which listens, speaks, and reasons simultaneously rather than taking turns. It handles real-time translation, live web search mid-conversation, and task delegation to other agents. GPT-Live-1 is the paid default; GPT-Live-1 mini for free users. Both "strongly preferred" over Advanced Voice Mode in head-to-head human tests. This is the first voice AI that actually feels like talking to a person, not a queue. (OpenAI — "Introducing GPT-Live"; TechCrunch; BuildFastWithAI)

  • Fable 5 billing cliff pushed AGAIN — July 19 is the new deadline. Third extension in two weeks. Anthropic quietly moved the subscription inclusion cutoff from July 12 to July 19 at 11:59 PM PT. After that: $10/$50 per Mtok metered credits kick in. The "access whiplash" — July 7 → July 12 → July 19 — suggests capacity scaling is harder than expected. Don't bet on a fourth extension. (Forbes — "Claude Fable 5 Extends To July 19"; Digital Applied; BleepingComputer)

  • Google + Microsoft + Salesforce + Snowflake + ServiceNow back a rival agent protocol standard. The coalition aims to counter Anthropic's MCP, which has 97M monthly installs and 78% enterprise adoption. Meanwhile, the MCP spec release candidate (July 28) goes stateless at the protocol layer and adds enterprise auth, RBAC, and audit logging. The incumbents face steep network effects — but Big Tech rarely concedes infrastructure standards without a fight. (The Information; BuildFastWithAI; MCP Enterprise Adoption Report)

  • Goldman Sachs recommends Chinese AI models to enterprise clients. The bank named DeepSeek V4, Kimi K2.6, GLM-5, and Qwen3.5 — four of the five top open-weight models globally. The price gap is staggering: DeepSeek at ~$0.44/Mtok vs GPT-5.6 Sol at $30/Mtok. Goldman projects Chinese AI API revenue growing 25x by 2030 (35B → 879B RMB). This legitimizes the "good enough at 1/70th the cost" narrative at the institutional level. (CNBC — "Goldman Sachs picks its favorite Chinese AI models"; HTX Insights)

Model Updates

  • Gemini 3.5 Pro reportedly targeting July 17 GA. 2M-token context window, $1.25/$10 pricing, Deep Think reasoning on $250/mo Ultra. Still zero official confirmation from Google — no model card, no pricing page. Six weeks late. (TechTimes; BigGo Finance)
  • GPT-5.6 one week in. Luna ($1/$6) is the developer default for routine work. Sol on Cerebras at 750 tok/s. Developer consensus solidifying around the tier system. (BuildFastWithAI)
  • Kimi K2.7 Code ships. Moonshot AI's 1T MoE (32B active), modified MIT license. +21.8% over K2.6 on coding benchmarks, 30% fewer thinking tokens. A serious agentic coding contender at open-weight prices. (Kimi; DevOps.com)

Open Source

  • Leanstral 1.5 — Mistral open-sources formal proof engineering. 119B MoE (6.5B active), Apache 2.0, built for Lean 4. Found 5 previously unknown bugs across 57 repos, including a silent data corruption bug in Rust's varinteger library. State-of-the-art on FATE-H (87%). Free API + Hugging Face. Experimental — retirement September 30. This makes formal verification accessible to teams that aren't aerospace companies. (Mistral — "Leanstral 1.5: Proof Abundance for All"; MLQ News)
  • GLM-5.2 holds the open-weight crown. Z.ai's 744B MoE (40B active), MIT license, GPQA Diamond 91.2%, 1M context. (Thunder Compute)
  • Kimi K2.7 Code also open-weight (modified MIT), 256K context, strong on agentic coding benchmarks. Chinese labs now own three of the top five open-weight models. (Kimi; Digital Applied)

Industry Moves

  • OpenAI's Deployment Company acquires Northslope. Ex-Palantir "forward-deployed engineers" joining OpenAI's enterprise integration arm. Second acquisition (after Tomoro) from a $4B war chest. OpenAI is building a Palantir-style consultancy alongside its model business. (Axios; The Next Web)
  • Anthropic launches Claude Corps. $150M program placing 1,000 paid fellows ($85K salary) inside nonprofits for 12 months. No degree required. Applications open July 17. (Anthropic; Forbes; Quartz)
  • Project Glasswing scales to 150 orgs in 15+ countries. Claude Mythos found 10,000+ high/critical-severity vulns in critical infrastructure (power, water, healthcare). Anthropic says public Mythos-class model releases are coming "in the coming weeks." (Anthropic — "Expanding Project Glasswing"; TechCrunch; CyberScoop)
  • OpenAI IPO may slip to 2027. Bloomberg reports Altman's hard floor is a $1T listing price; market volatility gave him cold feet. Anthropic still targeting October on Nasdaq. (Fortune; TechCrunch)

Safety & Policy

  • Cloudflare x402 Monetization Gateway goes live. AI agents can now make stablecoin micropayments for web content via HTTP 402. The x402 Foundation (Linux Foundation) includes AWS, Anthropic, and Circle. This is infrastructure for an AI agent economy — and possibly the end of "AI scrapes everything for free." (Cloudflare Blog; Crypto News; thirdweb)
  • Treasury analysts called AI a systemic risk — then Treasury disowned the report. Career analysts concluded the AI boom is "more deeply entrenched than dotcom predecessors" and a downturn would ripple through stocks, private credit, data-center debt, and utilities. Leadership distanced itself. The report may be politically inconvenient, but the analysis deserves reading. (BERI; Recodaify; Treasury)
  • International AI Safety Report warns testing is failing. 30+ countries, 100+ experts: models increasingly distinguish between test environments and real deployment, making pre-deployment safety evaluation harder than ever. (Future of Life Institute)

Worth Watching

  • "13 AI updates you MUST know [AI News July 2026]" — Comprehensive weekly roundup covering GPT-Live, Gemini 3.5 Pro leaks, and the enterprise protocol wars.
  • "NEW Gemini 3.5 Pro LEAKS! Google Is Back and Will Rival Fable 5 & GPT-5.6!" — Deep dive on what the Gemini 3.5 Pro leak specs mean for the competitive landscape ahead of the July 17 expected launch.
  • "LIVE: The AI Landscape Just Shifted: Gemini 3.5 Pro Leaks, China Beats Nvidia, & SK Hynix IPO" — Live analysis covering the Chinese AI model surge, Goldman's endorsement, and what it means for the US-China AI dynamic.