Introduction

Here are six key topics from overseas AI news collected in the latter half of July 12. From the battle over AI cost efficiency and debates over open-source philosophy, to new developments in a major lawsuit, the semiconductor supply chain outlook, geopolitical deregulation, and expanded developer tooling — it's a lineup packed with wide-ranging developments.

---

1. OpenAI, Meta, and SpaceX AI Roll Out New Models with 'Cost Efficiency' as the Main Battleground

According to Bloomberg, three major AI developers released new models over the past week. What's notable is that each model's biggest selling point is not 'what it can do' but 'how cheaply it can do it.'

OpenAI says its frontier model, GPT-5.6, can complete tasks with far fewer tokens than before. Tokens are the units of data AI models process, so lower consumption translates directly into cost savings. For enterprise customers, API costs are the top concern in production, which makes this improvement in 'token efficiency' hugely significant.

Meta and SpaceX's AI division are likewise putting cost front and center — a sign that competition among AI models is expanding from 'performance' into the dimension of 'economics.'

---

2. Zhipu Founder Says 'Frontier AI Should Be Available to Everyone,' Open-Sources GLM-5.2

Tang Jie, founder of Chinese AI lab Zhipu, has argued that frontier AI (cutting-edge AI) should not be monopolized by a handful of companies but made widely available. In an internal memo obtained by Bloomberg, he wrote that 'true safety comes not from technical barriers, but from broad participation, sharing, and oversight.'

In line with that stance, Zhipu has released its frontier model GLM-5.2 under an open-source license that allows commercial use. The debate over AI safety has long split between those who argue access should be restricted and those who believe openness is what makes AI safe — Zhipu's position represents the latter camp.

GLM-5.2 is already seeing active adoption across many developer communities, drawing attention as a sign of the open-source camp's growing influence.

---

3. Apple v. OpenAI Lawsuit: An Engineer's 'lol Moment' Reveals How the Improper Access Worked

New details have emerged in the case where Apple sued OpenAI for trade secret infringement. Apple argues that when former iPhone engineer Chang Liu moved to OpenAI's hardware division, what he took from Apple wasn't just experience.

According to the complaint, Liu kept a company-issued MacBook he was supposed to return and continued receiving inside information through his relationships with Apple employees. Most seriously, he allegedly exploited a software bug that enabled access to Apple's internal file servers, allowing him to keep tapping internal data even after his departure.

Apple accuses OpenAI's hardware chief of leading an organized campaign to poach information, and the looming courtroom battle will be closely watched. The case is emblematic of the rising tensions over intellectual property among big tech companies.

---

4. SK hynix CEO: Memory Shortage Could Persist Beyond 2030

Kwak Noh-Jung, CEO of SK hynix, one of the world's largest memory chipmakers, said the memory chip shortage driven by AI demand could continue beyond 2030. He made the remarks in his first English-language interview, coming just as the company completed the largest IPO by a Korean company in US history.

Customer companies are moving to lock in supply through long-term contracts, and the CEO noted that 'there is a view that the shortage will be prolonged.' Training and inference for AI models require vast amounts of HBM (high-bandwidth memory), and observers point to a structural problem: demand growth is outpacing the expansion of supply capacity.

---

5. US Eases AI Chip Export Restrictions on the UAE

The Trump administration has eased export restrictions on the UAE (United Arab Emirates), opening the door to exports of a broad range of technologies including advanced semiconductors. In a notice issued by the Commerce Department's Bureau of Industry and Security (BIS), the reasons cited include the steps the UAE has taken to protect American sensitive technologies, as well as its support for the United States during the war with Iran.

The UAE has been investing enormous sums in AI development, making this easing of restrictions major news for AI chipmakers like NVIDIA as well. At the same time, some voices warn of technology leakage via China, so the geopolitical balance will be worth watching.

---

6. Anthropic Extends Free Fable 5 Access and 50% Higher Claude Code Usage Limits

Anthropic has announced that it is extending, through July 19 US Pacific Time, both its campaign offering paid-plan users access to its latest model, Claude Fable 5, at no extra cost and its 50% increase in the weekly usage limits for the AI coding tool Claude Code.

Free access to Fable 5 had been scheduled to end on July 12, but was extended by another week just before the deadline — a sign that Anthropic is pushing hard to boost usage in a market where cost competition is intensifying.

---

Wrap-Up

Today's news shows the AI industry entering a new stage of competition, shifting from 'performance' to 'efficiency and economics.' OpenAI's token efficiency gains, Zhipu's open-source strategy, and Anthropic's expanded usage limits each take a different approach, but they all point in the same direction: delivering AI to more users, more cheaply. At the same time, as the Apple v. OpenAI lawsuit and the easing of restrictions on the UAE show, technological competition continues to spill over into courtrooms and international politics.