AI INTELLIGENCE / DAILY SIGNAL
0011011011
2026-05-29 / AI NEWS

AI News Intelligence Digest — May 29, 2026

Illinois passes first-in-nation AI safety audit law. SB 315 cleared the Illinois House 110-0 on May 27 and heads to Governor Pritzker, who has pledged to sign. The Artificial Intelligence Safety Measures Act requires fro...

BRIEFING STRUCTURE23 SIGNALS

Inside this digest

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

Top Stories

  • Illinois passes first-in-nation AI safety audit law. SB 315 cleared the Illinois House 110-0 on May 27 and heads to Governor Pritzker, who has pledged to sign. The Artificial Intelligence Safety Measures Act requires frontier AI developers (>$500M annual revenue) to undergo mandatory third-party safety audits, publish risk frameworks, and report AI safety incidents within 72 hours. Takes effect Jan 1, 2028. Both OpenAI and Anthropic praised it — a rare bipartisan, industry-supported AI regulation. This is the template other states will copy. (NBC News — "Illinois Legislature passes historic AI bill"; CBS Chicago; Crain's Chicago Business)

  • California's 30 AI bills survive crossover — the regulatory avalanche is real. Nearly all of California's AI bills cleared their chamber of origin before today's May 29 deadline. Highlights: AB 1988 (PAUSE Act on chatbot safety), AB 2023 (companion chatbot child safety), SB 813 (California AI Standards and Safety Commission), SB 947 (worker protections from AI), SB 951 (90-day notice for AI workforce displacement). Bills now cross to the second house June 1 with a July 2 summer target. Between Illinois, Connecticut (SB5 signed), and California, state-level AI regulation is moving faster than federal. (Transparency Coalition — "AI Legislative Update: May 29, 2026")

  • Computex 2026 kicks off — Nvidia teases "surprise product" alongside Vera Rubin. Jensen Huang's keynote is Monday June 1. Vera Rubin NVL72 already won multiple Best Choice Awards — a ~2 million part supercomputer with 150 Taiwan ecosystem partners. Huang teased: "We have a surprise new product that we haven't told anyone about yet." The Vera Rubin launch is being called "the largest product launch, probably in the history of Taiwan." Intel Arc G3 handhelds and budget laptops fighting MacBook Neo also expected. GTC Taipei is embedded directly into Computex this year, June 1-4. (Bloomberg; NVIDIA Blog; Tom's Guide)

  • Gemini Spark rolls out to US subscribers — Google's 24/7 cloud agent goes live. Google's always-on AI agent, announced at I/O on May 19, is now available to AI Ultra subscribers ($100/mo, down from $250). Spark runs in the cloud even when your device is off — drafting emails, monitoring inboxes, tracking deadlines across Gmail, Docs, Sheets, and third-party apps via MCP (Canva, OpenTable, Instacart). Early reports: it struggles with relationship context ("called a live-in boyfriend a close friend") but the ambient agent paradigm is here. (9to5Google — "Gemini Spark rolls out to Google AI Ultra in the US"; CNBC; TechCrunch)

  • Trump's AI policy vacuum deepens as Mythos fears persist. Developing story: Trump scrapped his AI executive order last week after billionaire lobbying (covered May 28). The broader context is now clearer — the proposed EO was voluntary, asking labs to share frontier models with government 90 days pre-release. Even that was too much for Sacks/Musk/Zuckerberg. Meanwhile, CAISI has quietly partnered with Google, Microsoft, and xAI for government evaluations anyway. The irony: Trump entered office opposing Biden's AI safety framework, and Mythos's cyber capabilities have pushed him toward similar policies. Pentagon still pushing for stricter rules. (Fortune; NBC News; Tom's Hardware)

Model Updates

  • GPT-5.5 Instant gets readability improvements; two older models discontinued. Canvas feature moved to direct chat. (OpenAI release notes)
  • Gemini 3.5 Pro confirmed for June release. Improved reasoning and long-context over Flash. Flash already GA at $1.50/$9 per 1M tokens with 76.2% Terminal-Bench 2.1. (Google I/O)
  • Grok V9 finished training (1.5T params, fed Cursor programming data). Public release in 2-3 weeks. SpaceX IPO June 12 at $1.75T — timing is deliberate. $60B Cursor acquisition closes ~30 days post-IPO. (Bloomberg; Basenor)
  • June model collision still on track: Claude 4.8, GPT-5.6, Gemini 3.5 Pro, MiniMax M3 all expected. DeepSeek V4 Pro's permanent pricing ($0.44/$0.87/1M tokens) continues to undercut everyone 10-30x.

Open Source

  • Quiet period for major open-source releases. ZAYA1-8B (Zyphra, Apache 2.0, MoE reasoning) and SubQ 1M-Preview (first commercial subquadratic LLM, 12M context) from earlier May remain the standouts.
  • Nvidia unveiled new open models, data, and tools at pre-Computex events to "accelerate AI across every industry." Full details expected at Jensen's June 1 keynote.

Industry Moves

  • SpaceX/xAI + Cursor: $60B acquisition option confirmed by Bloomberg. Deal closes ~30 days after June 12 IPO. Alternative: $10B partnership fee if acquisition not exercised. Cursor hit $3B ARR in April, projected $6B by year-end. The Cursor data pipeline feeds Grok V9's code capabilities. (Bloomberg; Entrepreneurloop)
  • Cohere + Aleph Alpha merger: Creates "transatlantic AI powerhouse" valued at $20B. Schwarz Group (German retail) backing with ~$600M. Focused on sovereign AI for regulated sectors (defense, finance, public sector). Both Canadian and German governments endorsed the deal. (TechCrunch; CNBC)
  • Four-acquisition week recap (from May 27 digest, now confirmed): Anthropic bought Stainless ($300M+), Mistral bought Emmi AI (physics-aware models), Google DeepMind hired Contextual AI team ($80-90M licensing), Meta acqui-hired Dreamer team. The frontier labs are buying capabilities rather than building. (StartupHub.ai)
  • Microsoft shipped computer-using agents to production in Copilot Studio — screen navigation, form-filling, sub-500ms voice latency. Meanwhile still unwinding Claude Code licenses (June 30 deadline). (Buildfastwithai)

Safety & Policy

  • Illinois SB 315 — first US law mandating third-party safety audits for frontier AI developers. 110-0 House vote. Takes effect Jan 2028. (See Top Stories)
  • California — 30 AI bills survive crossover deadline. SB 813 would establish a dedicated AI Standards and Safety Commission. (See Top Stories)
  • Connecticut SB5 signed — joins NY, CA, WA, OR, ID, IA with laws on harmful chatbot interactions.
  • EU AI Act amendments — Council and Parliament reached provisional agreement to streamline rules, extend compliance deadlines for high-risk AI systems (August 2 deadline), and add new rules on AI-generated intimate content. (EU Council)
  • YouTube auto-labeling now rolling out — no longer relying on creator self-disclosure. Internal detection flags "significant photorealistic AI" content. Labels move to prominent positions below video player (long-form) and as overlays (Shorts). (TechCrunch; Variety)

Worth Watching

  • "Illinois Legislature passes historic AI bill that would require third-party safety audits" — NBC News coverage of SB 315, the most consequential state AI law yet. https://www.youtube.com/results?search_query=Illinois+SB+315+AI+safety+bill
  • PBS News Hour — May 29, 2026 — Full episode with AI policy coverage amid the California crossover deadline and Computex preview.
  • Daily AI News from GAI Insights — 5/29/2026 — Expert analysts' daily morning briefing covering today's developments.