AI Business | Jul 9, 2026

Analysis

• Illinois became the first US state to mandate independent safety audits for frontier AI developers the same week a federal alternative drew criticism for overreaching the other way. Governor JB Pritzker signed the Artificial Intelligence Safety Measures Act, requiring annual third-party audits for labs above $500 million in revenue, while critics of the competing federal Great American AI Act called its sweeping preemption of state AI laws "net-negative as written."
• Microsoft advanced its AI-deployment strategy on two fronts in the same window: internally, it began substituting in-house models for OpenAI and Anthropic in Excel and Outlook to cut costs, while externally it deepened a $15.2 billion commitment to the UAE, where its Copilot software already runs on 35,000 government workstations.
• xAI and Google pulled in opposite directions on safety this window. A lawsuit alleges Grok let a user generate roughly 7,000 child-abuse images while xAI obstructed a law-enforcement investigation; separately, Google's SynthID watermark correctly debunked a viral AI-generated hoax image of Senator Mitch McConnell — a rare validated win for anti-deepfake technology, per TechCrunch.
• Frontier-model pricing diverged sharply this window. SpaceXAI's new Grok 4.5 launched at $2/$6 per million input/output tokens, undercutting Anthropic's Opus 4.8 ($5/$25) and OpenAI's GPT-5.6 Luna ($1/$6), even as OpenAI cleared US government review to publicly release the GPT-5.6 family on Thursday.
• Samsung's record $58 billion quarterly profit landed the same window SK Hynix moved to close its $28 billion Nasdaq ADR book and CXMT readied a $4.3 billion China IPO — yet KOSPI fell nearly 5% on the news, a sign investors see the AI-memory earnings run as already priced in.
• Dated catalysts inside 72 hours: OpenAI's GPT-5.6 public release Thursday, Jul 9; SK Hynix ADR pricing guidance due after Thursday's Korea market close.

POLICY & REGULATION

• LEGPOLITICO | Pritzker draws the AI line | Illinois Governor JB Pritzker signed the Artificial Intelligence Safety Measures Act, making Illinois the first US state to require the largest AI developers — frontier labs with more than $500 million in revenue — to create risk-mitigation frameworks and undergo annual independent audits disclosing how they assess catastrophic risks. The law builds on similar measures in California and New York but goes further by mandating third-party evaluations. "Where the federal government has been unwilling to step up, states must venture once more unto the breach," Pritzker said.
• LEGLawfare-the-great-american-ai-act) | Congress Should Do Something: The Case for (Fixing) the Great American AI Act-the-great-american-ai-act) | Lawfare argued the proposed federal Great American AI Act (GAAIA) is the strongest federal frontier-AI safety framework yet drafted but called it "net-negative as written" because its sweeping preemption would override state AI laws. The piece traces the shift to the White House's June 12 decision to invoke export-control authority barring Anthropic from letting foreign nationals access its Mythos-class Fable 5 model over cyber-capability concerns, after which senior administration officials floated an FDA-style mandatory predeployment licensing regime for frontier models.

PARTNERSHIPS & ENTERPRISE

• ENTBloomberg Business | Microsoft Replaces OpenAI, Anthropic With Own AI in Some Apps | Microsoft has begun replacing OpenAI and Anthropic models with its own in Excel and Outlook, citing cost reduction. Tens of thousands of AI prompts in the two applications are now completed weekly with Microsoft's in-house models rather than its partners'.
• ALLYSemafor | Exclusive: Microsoft deepens AI ties with UAE to automate government work | The United Arab Emirates is deepening ties with Microsoft as the government pushes to automate at least half of public-sector work with AI agents. Copilot already runs on 35,000 Abu Dhabi government employees' computers, and Inception42, a subsidiary of Abu Dhabi AI conglomerate G42, is building agents on the Microsoft platform to handle procurement tasks. Microsoft has committed $15.2 billion to the UAE through 2030 and holds a stake in G42; president Brad Smith has called the country a gateway for regional computing power.
• ALLYWired | OpenAI's Chief Futurist Is Leaving the Company | OpenAI chief futurist Joshua Achiam told colleagues he is leaving the company later this month after nearly nine years, saying his departure was not driven by any single reason. Achiam, who previously led the company's mission-alignment team and worked alongside global affairs chief Chris Lehane on AI-policy advocacy, said "the world is in on the secret now and it feels possible to work on the mission from outside the walls of a frontier lab." OpenAI has not named a successor for the role, which sat at the intersection of its safety and policy teams.

GOVERNANCE & SAFETY

• INCIDArs Technica | Lawsuit: Grok user made 7K child sex images; xAI only reported one gang rape prompt | An amended complaint in a proposed class action, expanded Tuesday, alleges xAI's Grok chatbot let a user generate roughly 7,000 child sexual abuse images from a single photo of his stepdaughter, taken when she was 11. The complaint alleges Grok's safety system flagged only one prompt — for "gang rape" — triggering a CyberTip to the National Center for Missing and Exploited Children, and that xAI then repeatedly declined to help law enforcement identify the user despite mandatory reporting requirements. The man was later arrested after police obtained a warrant; a forensic review found the 7,000 images, the complaint says.
• GOVTechCrunch | Google's deepfake detector system used to debunk McConnell hoax pic | Google's SynthID watermarking system was used to debunk a viral AI-generated image purporting to show Kentucky Senator Mitch McConnell in a hospital bed, which fact-checking site Snopes confirmed contained the SynthID signature. Launched at Google's I/O conference in 2025, SynthID embeds an invisible signature that survives screenshots; Gemini has carried the watermark since launch, OpenAI joined the program in May 2026, and Anthropic does not participate.
• SAFEAxios | AI learned faster than the tests designed to measure it | Federal agencies face an Aug. 1 deadline to establish a classified process for benchmarking frontier AI models' hacking capability, though the Financial Times reports standards could arrive as soon as this week. When Anthropic's Fable 5 model returned to service last week, the company said it was building a standardized jailbreak-outcome benchmark with Amazon, Google, and Microsoft. Testing lab Irregular, which works with OpenAI and Anthropic, released a new cyber benchmark in late June measuring tasks such as remote code execution and privilege escalation; Stanford's 2026 AI Index warned that "evaluations intended to be challenging for years are saturated in months."

RESEARCH & MODELS

• MODELUPDATEBloomberg Business | OpenAI to Roll Out Top AI Model Globally After Limited Preview | OpenAI will publicly release its GPT-5.6 Sol, Terra, and Luna models on Thursday, roughly two weeks after limiting the rollout to a "small group of trusted partners" at the request of the US government. The US Department of Commerce has cleared the broader release, Axios reported, and CEO Sam Altman wrote "Happy building" on X to confirm the launch. The public rollout follows Anthropic's restoration of its Claude Fable 5 and Mythos 5 models after Commerce lifted an export-control directive that had forced their suspension.
• MODELAxios | Scoop: SpaceXAI launches new model, Grok 4.5 | SpaceXAI released Grok 4.5, pitched as a coding and agentic-work model trained alongside Cursor that the company says outperforms comparable models on engineering and knowledge tasks. Elon Musk called it "an Opus-class model, but faster, more token-efficient and lower cost," pricing it at $2 per million input tokens and $6 per million output tokens versus Anthropic's Opus 4.8 at $5/$25 and OpenAI's GPT-5.6 Luna at $1/$6. Grok 4.5 was trained on the same compute capacity SpaceXAI leases to competitors Anthropic and Google, and is not yet available in the EU.
• SCIThe Economist | A new era of biology beckons | Francis Crick Institute director Edith Heard, who helped Google DeepMind build its Nobel Prize-winning AlphaFold protein-structure model, discussed AI's expanding role in biology research in an Economist podcast interview.
• MODELBloomberg Business | Tencent WeChat AI Agent Shows Promise in Super-App Fight: Review | Tencent Holdings is testing Xiaowei, a prototype AI agent inside WeChat designed to eventually run errands for users, as the company works to restore its reputation as a frontier-AI innovator after falling behind rivals in model development.

COMPUTING & INFRASTRUCTURE

• CHIPUPDATEThe New York Times | Samsung Made More Profit Last Quarter Than the Last Two Years Combined | Samsung Electronics reported preliminary April–June operating profit of 89.4 trillion won (roughly $58 billion), nearly 20 times higher than a year earlier and more than 2024 and 2025 combined, on demand for memory chips used in AI data centers. Revenue more than doubled to about $112 billion. Despite the results, Samsung shares fell more than 8% in Seoul, and the broader KOSPI index tumbled nearly 5% — triggering circuit breakers for the 12th time in its history — as Goldman Sachs analysts warned of a "bumpy path" after gains they said were 90% attributable to Samsung and SK Hynix.
• CHIPBloomberg Business | Samsung Mass Produces Storage Drives for Nvidia's Vera Rubin | Samsung Electronics began mass production of its PM1763 enterprise solid-state drive, its most advanced data-center storage product, for use in Nvidia's upcoming Vera Rubin platform, first unveiled at Nvidia's GTC conference earlier this year.
• CHIPBloomberg Business | Chinese AI Startup DeepSeek Developing Own AI Chip, Report Says | DeepSeek is developing its own AI accelerator and taking direct control of its hardware roadmap, according to separate reports from Bloomberg and Semafor, as the Chinese lab pursues compute self-sufficiency amid US export controls on advanced chips.
• CHIPReuters | Synopsys to cut chip fab manufacturing control software in shift to AI design, sources say | Synopsys is discontinuing a suite of chip-fabrication process-control software, including its Equipment Engineering System and Fault Detection and Classification tools, to redirect resources toward higher-margin AI design offerings, six sources told Reuters. The company notified more than 10 chipmakers, including Samsung Electronics, SK Hynix, and Kioxia Holdings, and has already laid off dozens of staff.
• CHIPBloomberg Business | Semiconductor Worker Shortfall Endangers US Chip Factory Revival | A growing shortage of high-skilled semiconductor workers threatens to delay construction of new US chip plants and constrain future production, according to a new industry report, unless companies pool resources and government funding continues.
• CHIPAxios | The AI chip rush is crowding the same narrow pipeline | Axios reported that the surge in custom AI chip development from Nvidia rivals is straining the same limited pool of advanced semiconductor manufacturing and packaging capacity.
• CHIPBloomberg Business | Nvidia's $1 Trillion Slide Sends Valuation to Pre-AI Boom Levels | Nvidia has lost roughly $1 trillion in market value in under two months, leaving the stock at its cheapest valuation since before the AI boom, even as its GPUs continue to dominate AI data-center demand.
• CHIPTechCrunch | Hot French startup ZML releases free product to speed inference across lots of AI chips | French startup ZML released a free product designed to speed AI inference across multiple types of AI chips, per TechCrunch, aiming to reduce hardware lock-in for companies running inference workloads.

Still in play:

• South Korea's AI chip boom is concentrating gains among Samsung and SK Hynix, widening the gap with smaller domestic suppliers, according to the Financial Times. [Financial Times]

GENERATIVE MEDIA & EMBODIED AI

• BOTBloomberg Business | Mistral AI Releases Robotics Model to Support Physical AI Push | Mistral AI released Robostral Navigate, a robotics navigation model that lets robots move through complex environments, as the French startup expands into physical AI after signing deals with major European industrial customers.
• GENThe New York Times | Meta Unveils an A.I. Image Generator | Meta released Muse Image, an AI image generator available on Instagram, WhatsApp, and the Meta AI app, in its latest push to catch up with rivals Google, Anthropic, and OpenAI. Originally codenamed Mango, the tool is the second major release from Meta Superintelligence Labs, led by Alexandr Wang, following April's Muse Spark chatbot model. Muse Image lets users edit generated images via sketches or annotations and will power advertiser-specific tools inside Meta's Advantage Plus ad platform; power users must subscribe to a Meta One plan for higher usage limits. Meta also previewed an AI video generator, Muse Video, due within "the coming months."

FUNDING & DEALS

• IPOUPDATEReuters | SK Hynix to close $28 billion ADR bookbuild on Wednesday after oversubscription, source says | SK Hynix will close bookbuilding on Wednesday US time for its $28 billion American depositary receipt offering — one of the largest new share sales ever — after orders covered the books multiple times over, a source told Reuters. Pricing guidance is due after Thursday's close of the South Korean market.
• IPOUPDATEBloomberg Business | Memory Chipmaker CXMT to Launch $4.3 Billion China IPO Next Week | Memory chipmaker CXMT Corp., a Samsung Electronics and SK Hynix competitor, will open investor subscriptions next week for a $4.3 billion public offering of 6.688 billion shares — the most anticipated China listing of the year.
• RNDTechCrunch | AI chip maker SambaNova raises $1B at $11B valuation, 5 months after last mega round | AI chip maker SambaNova raised $1 billion at an $11 billion valuation, just five months after its last mega-round, as investors continue backing challengers to Nvidia. JPMorgan Chase said it would deploy SambaNova's systems for on-premises AI inference, and rival Korean chip startup Rebellions told CNBC it is preparing an IPO on the Kospi for 2027.
• RNDBloomberg Business | Nvidia Rival Positron Holds Talks to Raise Funds at $5 Billion Valuation | AI chip startup Positron is in talks to raise about $750 million in a two-tranche round, with the first tranche valuing the company at $3.5 billion, as it seeks to compete with Nvidia.

Calendar

• MODELJul 9 (Thu) — OpenAI to publicly release its GPT-5.6 Sol, Terra, and Luna models globally.
• IPOJul 9 (Thu) — SK Hynix ADR pricing guidance due after Thursday's Korea market close.

MARKETS

8 Jul 2026 close | Retrieved 8 Jul 21:45 UTC | Yahoo Finance

AI Equities (1D) | Nvidia 204.12 USD +3.7% | Microsoft 383.34 USD -1.4% | Alphabet 361.92 USD -1.4% | Meta 603 USD -2.0% | Amazon 243.62 USD -1.0% | Palantir 132.22 USD -1.6%

Semiconductors (1D) | AMD 517 USD +0.3% | TSMC 436.98 USD +1.0% | Broadcom 388.69 USD +4.8% | ARM 300.24 USD -0.1% | Super Micro 28.17 USD +7.3%

AI Infrastructure (1D) | CoreWeave 90.00 USD +7.7%

Indices (1D) | NASDAQ 25,871 +0.2% | SOX 12,575 +2.2%

Coverage: 7 Jul 03:15 – 9 Jul 03:15 UTC