Good morning. It's Monday, July 13, and we're covering foreign influence campaigns targeting AI infrastructure, SK Hynix's record-breaking IPO, and growing scrutiny over Chinese access to U.S. AI models.

Plus: this week's prompt to never forget a name again, and our latest Zeitgeist discussion on the economics reshaping the AI race.

Apple filed a complaint in federal court alleging OpenAI systematically stole trade secrets to build OpenAI's forthcoming consumer hardware, singling out OpenAI Chief Hardware Officer Tang Tan (a former Apple VP) for allegedly directing job candidates to bring Apple parts to their interviews for show-and-tell sessions and circulating an internal Apple offboarding document to teach new hires how to evade security checks. The suit also names former Apple engineer Chang Liu, who allegedly kept his work laptop after leaving, exploited a bug to access Apple's network storage, and messaged a former Apple colleague, "LOL, I found out I can access the [network storage], so funny.”

Despite their parent companies appearing on the Pentagon's 1260H blacklist, Singapore subsidiaries of Alibaba, Baidu, and Tencent legally accessed AI services from OpenAI and Google under current U.S. rules. The disclosures have renewed pressure in Washington to expand AI export controls beyond semiconductors. OpenAI recently cut Alibaba-linked API access over suspected model distillation, while Anthropic bars Chinese firms and their overseas affiliates entirely.

Cursor is quietly building a general-purpose AI assistant, signaling its next growth target is office workers rather than developers. The internal project, called Sand, aims to handle emails, spreadsheets, and everyday business tasks while competing with Anthropic's Claude Cowork and OpenAI's ChatGPT Work. Development began after Cursor secured SpaceXAI compute in April and comes ahead of SpaceX's planned $60 billion acquisition, though the product has not been approved for launch.

Rather than proposing a single AI law, Senator Ed Markey has unveiled a broad legislative package spanning datacenters, hiring algorithms, healthcare, and child safety. The agenda would require federal certification for AI datacenters, restrict employers from relying primarily on automated hiring decisions, and mandate bias audits for high-impact AI systems. Additional measures would strengthen protections for children, require human overrides in healthcare, and standardize reporting on AI's energy and environmental impact.

Grok 4.5 tops agentic coding benchmarks, beating Opus 4.8, while xAI trains its next model using Cursor data and Composer 3 nears launch.

SK Hynix raised $26.5 billion through its Nasdaq debut on July 10, 2026, making it the largest U.S. IPO ever by a non-American company and surpassing Alibaba’s 2014 record. The South Korean memory chip maker sold 177.9 million American depositary shares at $149 each, with demand reportedly exceeding available shares by more than seven times as investors bet on continued AI infrastructure growth.

The company plans to use the proceeds to expand memory chip production in South Korea, including a new fabrication plant, advanced packaging facilities, and extreme ultraviolet (EUV) lithography equipment. The debut comes as U.S. Commerce Secretary Howard Lutnick urges SK Hynix and Samsung to build new U.S. chip factories, underscoring growing competition between Washington and Seoul over where the next generation of AI semiconductor manufacturing will be based. → Continue reading here.

China, Russia, and Iran are increasingly using state media and coordinated online campaigns to amplify U.S. opposition to AI data centers, according to a new analysis by threat intelligence firm Alethea. The efforts include AI-generated social media posts, propaganda articles, and videos highlighting concerns over electricity costs, environmental impacts, and local infrastructure, while seeking to deepen existing political divisions rather than create new ones.

OpenAI said it identified a small China-linked operation that used ChatGPT to generate some of the content, although it found little authentic engagement before the accounts were removed. The findings come as AI infrastructure becomes a strategic priority for Washington, with policymakers warning that foreign adversaries are attempting to influence public opinion as the U.S. races to expand computing capacity. → Read the full article here. (Paywall)

The fight over AI's potential harms has played out state by state; Massachusetts Democrat Ed Markey, 79 and running for a third full term, wants to make it federal in one go. His "AI Accountability Agenda: Taking Power Back from Big Tech," rolled out Friday as a Guardian exclusive, bundles close to a dozen bills spanning workplace surveillance, biased algorithms, environmental impact, and kids-and-chatbots. The centerpiece is the boldest: a preliminary bill that would force any company that wants to build a data center to get FCC certification affirming it "will not harm the public interest" before a shovel goes in.

Congress has favored speed over guardrails, most of Markey's AI bills are stuck, and the White House just told the national security apparatus to accelerate AI adoption. The pieces actually moving are narrower: a Senate kids' online safety bill in March, the Hawley-Moore GUARD Act pushing to keep minors off AI companions, Illinois becoming the first state to require third-party frontier-model audits, and a 269-page House draft (the Great American AI Act) that would preempt state AI laws for three years. Markey traces the whole fight back to his father, who lost a finger to a factory machine before workplace safety laws existed, and argues AI is the newest version of the same gap: technology keeps outrunning the rules meant to contain it.→ Read the full article here.

Every morning the Forward Future team argues about the stories, papers, and launches that matter. What's real, what's hype, and where AI is headed next. These are the public notes.

Is AI's next arms race about intelligence, or about who can run it cheapest? Friday's team broke down a new "vibes" benchmark that has Claude losing ground on price, debated whether soaring compute costs mean a bubble or just scarcity, and asked why Anthropic just put Ben Bernanke on its oversight board.

▸ Why a new benchmark has the team questioning Claude's price-to-performance
▸ Why rising AI prices signal a compute shortage, not a bubble
▸ Why Anthropic brought in a former Fed chair with no equity in the company

Read Friday's Zeitgeist →

Musk Reverses on Anthropic: Elon Musk called Anthropic the AI leader after partnering with SpaceX, reversing months of public criticism.

Meta Rallies on AI Push: Meta shares surged 15% this week as investors cheered new AI models, custom chips, and a clearer strategy beyond ads.

Microsoft Expands Foundry AI: Microsoft added GPT-5.6 models, hosted agents, and an Asia-Pacific data zone to scale enterprise AI.

OpenAI Bets on Families: OpenAI is hiring a families-focused PM as ChatGPT's audience skews older.

Meta AI Detector Falls Short: Reuters found Meta's preview AI detector missed 55% of cropped AI images, exposing limits of watermark-based detection.

Nobel Chemist Joins China (Paywall): Nobel winner Omar Yaghi left UC Berkeley to lead a Tsinghua AI institute focused on discovering new materials.

Thanks for reading. See you next time!

— Matthew Berman, Nick Wentz & the Forward Future Team