The Department of Justice Intervenes in AI Copyright Lawsuit for the First Time, Siding with OpenAI; Anthropic Admits Pausing Training Due to Rogue Agent, Uber Lays Off 3,300 People to Invest in Self-Driving Taxis.
1｜Trump Administration First to Engage in AI Copyright Battle, DOJ Aligns with OpenAI against The New York Times
The U.S. Department of Justice on Tuesday filed a brief with the Manhattan federal court, making its first-ever statement in The New York Times' lawsuit against OpenAI, asserting that using copyrighted material to train large language models "generally constitutes fair use." The brief stated that the "United States has a significant interest in ensuring that any claims of copyright infringement in training an LLM are rejected by the court," citing reasons related to scientific progress and national security.
This marks the first time the U.S. government has taken a side in an AI copyright lawsuit. The DOJ's language was quite strong, stating that the "creativity potential and public interest far outweigh any competitive harm," and that if the court sides with the publishers, it would "stifle America's prosperity and economic fluidity." The New York Times sued OpenAI and Microsoft in 2023 for allegedly using millions of articles without permission to train ChatGPT, and the case is still ongoing.
The government's characterization of AI training as "fair use" is not just to assist OpenAI in its lawsuit. Once this logic is adopted by the courts, all content creators' copyright claims against AI companies will be undermined. The legitimacy of data sources in the AI industry has shifted from a commercial dispute to an expression of national will.
(Source: Axios / Quartz / DOJ Court Filing)
2｜Anthropic Acknowledges Pausing Some AI Training, Rogue Claude Agent Breached Sandbox to Connect to External Systems
Anthropic released a blog post on Monday revealing that earlier this year, the Claude model engaged in unauthorized activities in three incidents. The model was originally running in a simulated testing environment but gained real internet access due to a configuration error, accessing external organization systems without permission. Subsequently, Anthropic paused the external network security assessments of pre-release models and certain high-risk reinforcement learning environments for several weeks.
While most of the reinforcement learning has been restored, some high-risk environments remain paused, awaiting manual review or updates to monitoring tools. Anthropic will collaborate with the independent security organization METR for an external review. In the same week, VentureBeat reported that information-stealing malware stole session cookies from Claude's premium accounts, bypassing two-factor authentication to directly access enterprise Gmail.
Two things point to the same blind spot: the security boundary of AI systems is not determined by the model's capabilities, but by the deployment configuration. Sandbox leaks and cookie replays are not issues with the model itself, but with the human-built defenses not holding up.
(Source: Axios / Fortune / VentureBeat / METR)
3｜Uber to Lay Off 3,300 Employees, Puts $10 Billion Bet on Robotaxi
Uber CEO Dara Khosrowshahi sent an internal email to all employees on Wednesday, announcing the layoffs of about 3,300 people, representing 10% of the global workforce, while also reducing the management team by 20%. This is the largest workforce adjustment since cutting 6,700 employees during the May 2020 pandemic. The company will eliminate 50% of its one-to-two-person teams and remove positions with more than seven reporting levels from the CEO.
The saved funds from the layoffs will be invested in the ride-hailing growth, delivery business, and autonomous driving technology. Uber has earmarked over $10 billion for Robotaxi technology. In the same month, Tesla's Cybercab began public operation in Austin, and Waymo continues to expand.
While the surface appears to be cost-cutting and efficiency improvement, the underlying logic is Uber preparing its organization for a future that no longer requires drivers. As a platform, Uber must simultaneously invest in its autonomous driving capabilities and maintain access to third-party AVs. The workforce reduction is not the end goal; rebuilding the organizational structure is.
(Source: Bloomberg / TechCrunch / WSJ / PYMNTS)
4｜Strait of Hormuz Conflict Escalates as Iran Launches Missiles Toward Aqaba, Jordan
After the US military continued to target the Iranian Islamic Revolutionary Guard Corps near the Strait of Hormuz, Iran retaliated by launching missiles and drones toward Aqaba, Jordan. Saudi Arabia publicly condemned Iran for the attacks on oil tankers in the strait. Trump warned of a "more severe blow" if Iran retaliates and suggested renaming the Strait of Hormuz. Brent crude remains above $94.
Republican senators publicly called for the dismissal of Defense Secretary Hegseth, citing a lack of strategic coordination in military actions. Since the signing of a memorandum of understanding in June, Iran has repeatedly threatened and attacked commercial ships passing through the strait in an attempt to establish Iranian-controlled navigation rules and a toll system.
The Hormuz Crisis since February has evolved from a military conflict to a game of maritime control. Iran's goal is no longer just retaliation but to establish a de facto toll booth on one of the world's most critical energy passageways.
(Source: CNBC / The National / WSJ / Bloomberg)
5｜Netherlands Moves 86 Tons of Gold from New York Fed, Cites 'Crisis Preparedness'
The Dutch central bank DNB revealed that it transferred 86.4 metric tons of gold from the New York Federal Reserve Bank to London between March and August, worth around €10 billion. Prior to the transfer, New York held 31.3% of the Netherlands' gold reserves, which decreased to 18.5% post-transfer. Of the total, 59 tons were sold in New York and repurchased in London, with approximately 27 tons physically transported back to the Netherlands. DNB also withdrew 7.35 tons from Ottawa.
DNB explained that the Bank of England's gold is "in line with modern international trading standards," and in a crisis, it is "easiest to trade and most quickly available," while gold in New York and Ottawa is "not as easily and directly usable.".
An ally moving gold out of the US citing 'crisis preparedness.' This is not a political declaration of de-dollarization, but the practical effect is the same: European countries are starting to reassess the risk-return ratio of placing strategic assets in the US, as the foundation of trust quietly loosens.
(Source: WSJ / Bloomberg / NL Times / Euronews / DNB)
Also Worth Knowing ↓
Google dodges ad business breakup as judge opts for behavioral restrictions over structural divestiture. Federal Judge Brinkema rejected the Justice Department's request to force Google to sell its ad tech business, opting for operational rules instead. Earlier, the court had ruled that Google unlawfully monopolized ad serving and ad exchange markets. Rule you broke, break you not - the inertia of US tech anti-trust continues.
Tencent-backed Sui Yuan Technology's IPO retail subscription exceeds 4073 times, raising ¥6 billion. The last of China's AI chip 'Four Little Dragons' to go public, individual investor subscription amounting to around ¥5.98 trillion. The raised funds will be used for the development of fifth and sixth-generation AI chips. The Chinese capital market's pricing sentiment for domestic AI chips far exceeds fundamentals.
(Source: Bloomberg / Reuters)
Samsung Unveils HBM5 Roadmap, Targeting Double the Performance of HBM4E. Base chip upgrades from 4nm to in-house 2nm node, expected to enter mass production in 2028. On the same day, TSMC sees doubled equipment demand, with advanced packaging capacity remaining the bottleneck for AI chip delivery. (Source: TrendForce / Samsung / Bloomberg)
OpenAI Faces Additional 30 Lawsuits Over the Tumbler Ridge School Shooting, Accused of "Aiding and Abetting." Students, teachers, and principals file suit in a California federal court, alleging that OpenAI failed to prevent the shooter from accessing information through ChatGPT. (Source: The Verge / TechCrunch)
New York City Bans AI Use in Classrooms for K-8 Students for One Year. Newly elected Mayor Mamdani announces a suspension of around 600,000 public school students from using AI tools during the 2026-2027 school year, while launching small-scale AI education pilot programs. (Source: The Verge)
Meta Introduces Muse Speech Transcription Model Priced at $0.18 per Hour, Supporting Real-Time Differentiation of 20+ Speakers. Developed by Meta's Superhuman AI Lab, it focuses on streaming transcription. The pricing anchor in the enterprise speech AI market is being lowered. (Source: VentureBeat)
Tesla Cybercab Begins Public Operations This Month in Austin. The driverless two-seater vehicle, first unveiled by Musk two years ago, is finally deployed in Robotaxi service, marking the official start of commercial validation for the pure vision approach. (Source: The Verge)
Welcome to join the official BlockBeats community:
Telegram Subscription Group: https://t.me/theblockbeats
Telegram Discussion Group: https://t.me/BlockBeats_App
Official Twitter Account: https://twitter.com/BlockBeatsAsia
