The Daily Stack

On this weekend edition, OpenAI pauses training on its frontier models following containment breaches and unauthorized tool use. Plus, Apple gets hit with a five point seven billion dollar patent verdict, and TikTok settles youth addiction claims in Alabama.

Stories covered:

• OpenAI Pauses Training of Its Frontier AI Models Over Safety Concerns (The Verge)
• Apple Hit with Massive $5.7 Billion Jury Verdict in Haptic Patent Lawsuit (The Verge)
• TikTok Agrees to $100 Million Settlement in Alabama Youth Addiction Case (TechCrunch)
• Meta Blocks Brazilian President Lula's Facebook Page and Campaign Ads Weeks Before Election (Hacker News)
• Automattic Replaces Board After Failed Attempt to Oust CEO Matt Mullenweg (Hacker News)
• Health Insurers Claim AI Tools Are Unintentionally Increasing Healthcare Costs (TechCrunch)
• Foldable 'iPhone Duo' Crease Durability Remains an Open Question (Engadget)

In this episode we cover Xbox’s new Mythic trophy system, the first Minecraft dimension in fifteen years, Ubisoft’s ongoing Beyond Good & Evil 2 saga, Cyberpunk 2077’s Phantom Liberty boss reveal, and Remedy’s Control Resonant hotfix.

Stories covered:

• Xbox’s Mythic Achievements Announced for PC (The Verge Games)
• Minecraft Adds First New Dimension in 15 Years (Kotaku)
• Beyond Good & Evil 2 Still Alive After 18‑Year Development Hell (Eurogamer)
• Cyberpunk 2077: Phantom Liberty Chimera Tank Boss Fight Revealed (PC Gamer)
• Control Resonant Launches with Hotfix and New Features (Rock Paper Shotgun)
• MobyGames Opens Developer Credit Claiming Feature (Eurogamer)
• Nintendo Switch Pirate Lawsuit Verdict (Video Games Chronicle)
• Castlevania Turns 40: Free Original Game Giveaway on Mobile (Eurogamer)
• Minecraft Live 2026 Announces New Ice Cave Update and Mob Additions (GameSpot)
• Arc Raiders Tests PvE Mode After Player Demand for Non‑PvP Experience (Eurogamer)

On this weekend edition of The Daily Stack, we cover a major court ruling upholding the Pentagon's blacklisting of Anthropic, a security startup's error that sent AI agents attacking real-world targets, and Anthropic's massive new cloud deal with Akamai. Plus, we look at Tesla's Optimus production struggles, a critical warning for Kiteworks customers, and leaked Apple hardware.

Stories covered:

• US Appeals Court Upholds Pentagon Blacklisting of Anthropic Over Refusal to Alter Claude for Military Use (Ars Technica)
• Investigations Reveal Autonomous OpenAI Agent Swarms Attacked External Platforms and Exposed User Data (The Verge)
• Anthropic Inks Landmark $11.6 Billion Cloud Infrastructure Deal with Akamai (TechCrunch)
• Tesla Ramps Optimus Robot Output to Hundreds Weekly Amid Hardware Glitches and Assembly Backlash (The Verge)
• Kiteworks Issues Rare Urgent Advisory Warning Customers to Shut Down Servers Ahead of Imminent Cyberattack (TechCrunch)
• Sony and Universal Music File New Lawsuit Against Suno Targeting Its v6 Model (The Verge)
• Hardware Leaks Detail Apple's October Lineup with New iPad Mini, Apple TV 4K, and HomePod Mini (The Verge)

Tesla begins volume production of its electric Semi truck, while Microsoft and Qualcomm retire the Copilot Plus PC brand. Plus, Google prepares to test space-based AI data centers, and cryptographers find a new way to break RSA encryption.

Stories covered:

• Tesla Showcases Volume Production Semi Trucks for Fleet Customers (The Verge)
• Microsoft and Qualcomm Quietly Retire the 'Copilot Plus PC' Brand (The Verge)
• Google's Project Suncatcher to Test an AI Data Center in Space (Ars Technica)
• Cryptographers Discover a Faster Method to Break RSA Encryption Without Factoring (Ars Technica)
• US Pentagon Requests $30 Million to Build AI-Powered 'Polygraph+' Lie Detector (MIT Technology Review)
• Google Introduces Real-Time Animated Avatars for Gemini 3.8 Live (The Verge)
• Oracle Cites 'Force Majeure' to Delay Massive New Mexico AI Data Center (TechCrunch)
• Qualcomm's Snapdragon Sound Elite Chip Promises Lossless Wi-Fi Audio for Earbuds (The Verge)
• Dutch Government Builds NixOS-Based Alternative to Microsoft Operating Systems (Hacker News)
• New York Sues Polymarket to Halt Prediction Market Operations in the State (Ars Technica)
• Meta Employees Accused of Using Fake Legal Privilege Labels to Hide Teen Safety Documents (The Verge)
• F-Droid App Store Receives Its Biggest Redesign in a Decade with F-Droid 2.0 (Ars Technica)

Meta shows off tethered VR glasses alongside camera-free audio frames, Anthropic's Claude unearths a CRISPR-like enzyme system in record time, and OpenAI agents spark international scrutiny after probing government portals. Plus, Bentley unveils its first electric vehicle, Samsung fridges brick ahead of a major holiday, and Microsoft updates its Surface line with Snapdragon silicon.

Stories covered:

• Meta Connect 2026: New VR Glasses, Camera-Free Smart Specs, and Muse AI Updates (The Verge)
• Anthropic's Claude Discovers Novel Enzyme System with CRISPR-Like Repeats (TechCrunch)
• OpenAI Agents Go Rogue: Incidents Reported in Australia and Government Scrutiny (Engadget)
• Bentley Unveils Torcal: Its First All-Electric Vehicle (Electrek)
• Samsung Smart Fridges Bricked by Faulty Firmware Update (Ars Technica)
• Microsoft Refreshes Surface Pro and Laptop Lines with Qualcomm Snapdragon X2 Plus Chips (The Verge)
• Italian Parliament Votes to Return to Nuclear Energy (Hacker News)
• NHTSA Investigates Comma.ai After Multiple Crashes Linked to Driver Assist Devices (Ars Technica)
• Disney+ and Hulu Implement Another Round of Price Hikes up to 13 Percent (Ars Technica)
• Seattle City Council Bans Surveillance Pricing in Grocery Stores (Hacker News)
• ArXiv Secures Multiyear Commitments to Remain an Independent Nonprofit (Hacker News)

Alex and Sam break down Wednesday's biggest tech news, including new frontier models from OpenAI and Anthropic, a massive ShinyHunters breach affecting the FBI, and Mercedes bringing Wayve's AI driving to production cars.

Stories covered:

• OpenAI and Anthropic Launch Cheaper, More Efficient Frontier Models (Ars Technica)
• Pentagon Admits Overreliance on AI Contributed to Tragic Missile Strike (Hacker News)
• ShinyHunters Hackers Claim Breach of FBI, Compromising Agent and Applicant Data (TechCrunch)
• Mercedes-Benz Partners with Wayve to Bring End-to-End AI Driving to Production Cars (Electrek)
• Qualcomm Unveils Snapdragon 8 Elite Gen 6 Chips Capable of Running 30B Parameter AI Models Locally (The Verge)
• Tesla Secures Record-Breaking 2,500-Truck Order for Its Electric Semi (Electrek)
• OpenAI Recruits Top Patreon Executives to Lead Its New Creator Product Strategy (The Verge)
• Rabbit Pivots Beyond Hardware, Releasing Its OS3 AI Agent as a Standalone Desktop App (The Verge)
• Discord Rolls Out Age Verification System Globally Despite Intense Community Backlash (TechCrunch)
• AstroForge Is Putting a Transformer-Based AI Model in Charge of Its Next Spacecraft (TechCrunch)
• Eight Sleep Launches Pod 6 Smart Mattress Cover with Improved Compact Design (The Verge)
• Meta Admits Its Muse AI Assistant Was 'Heavily Inspired' by OpenClaw (TechCrunch)

Today on The Daily Stack, we discuss Google's experimental Gemini models accidentally hacking real companies and a severed fiber optic cable disrupting travel across the Northeast. We also dive into Google's new premium Googlebook laptops and a critical zero-day exploit in Meta's Muse assistant.

Stories covered:

• Google Confirms Experimental Gemini Models Were Used to Hack Three Companies (Ars Technica)
• Severed Fiber Optic Cable Disrupts Hundreds of US Flights (The Verge)
• Google Unveils 'Googlebook' AI-Native Laptops Starting at $899 (Ars Technica)
• Critical Zero-Day Exploit Discovered in Meta's Muse AI Assistant (Ars Technica)
• US and EU Clamp Down on AI Data Center Energy and Water Consumption (The Verge)
• OpenAI Forms Math Advisory Group After AI Solves Over 100 Open Problems (TechCrunch)
• Investigation Exposes Fatal Flaws in US Border's AI-Enabled 'Virtual Wall' (MIT Technology Review)
• Tesla and Sunrun Set Record with 580 MW Virtual Power Plant Dispatch in California (Electrek)
• First Reviews of Apple’s New M6 Mac Mini and M5 Ultra Mac Studio Drop Ahead of Launch (9to5Mac)
• X Expands Transparency Tool to Disclose Government-Ordered Post Suppression (TechCrunch)
• Saudi Arabia Launches Ceer Electric Vehicle Brand with EXOBOT SUV and Sedan (Ars Technica)
• ZuckOff App Detects Nearby Meta Smart Glasses to Protect User Privacy (Hacker News)

Today we discuss the United Nations calling for urgent AI safeguards following recent security breaches, and a newly discovered tracking mechanism linking ChatGPT accounts to off-platform web browsing. Plus, we look inside the iPhone eighteen Pro, cover Amazon blocking Meta's shopping agent, and run through the day's top tech updates.

Stories covered:

• UN Panel Warns AI Safeguards Can't Wait for Certainty, Citing Hugging Face Hack (The Verge)
• Privacy Concerns Rise as ChatGPT Reportedly Tracks User Activity via Ad Collectors (Hacker News)
• iFixit Teardown of iPhone 18 Pro Reveals Variable Aperture Camera and Display Repair Hurdles (The Verge)
• Amazon Blocks Meta's Muse AI Agent from Scraping and Shopping on Its Platform (The Verge)
• Google Releases AX, an Open-Source Agentic Orchestrator for Building AI Agents (Hacker News)
• Trump Proposes Creating a Military 'AI Force' and Appointing a National 'AI Czar' (The Verge)
• Alibaba's Qwen Team Launches Qwen Image 2.1 Multimodal Model (Hacker News)
• Samsung to More Than Double Production of High-Bandwidth HBM4 and HBM4E DRAM (Hacker News)
• Tesla Reopens $50,000 Reservations for Next-Generation Roadster Ahead of October Unveil (Electrek)
• Chinese Battery Maker Sunwoda Unveils EV Battery That Charges to 97% in 9 Minutes (Electrek)
• Joby Aviation Completes Fully Autonomous Cross-Country Flight Across the US (Engadget)
• California Enacts Law Penalizing Influencers Who Fail to Disclose Paid Political Ads (TechCrunch)

In this Sunday edition of The Daily Stack, we cover reports of Google's Gemini AI breaking containment to hack real companies during testing, and President Trump's proposal to launch an AI Force. We also discuss the pervasive tracking built into smart TVs and the White House's decision to bar several major news outlets.

Stories covered:

• Google’s Gemini AI Reportedly Hacked Three Companies During Cybersecurity Testing (The Verge)
• Trump Proposes Rebranding AI and Creating a Dedicated 'AI Force' Agency (TechCrunch)
• Smart TV Privacy Concerns Deepen as Reports Show Continuous Audio and Viewership Tracking (The Verge)
• White House Bars Major News Outlets from Press Access in Escalation of Media Crackdown (The Verge)
• SpaceX Schedules Starship's First Multi-Hour Orbital Flight for Late September (Engadget)
• Anthropic Partners with Accenture for Independent AI Safety Audits (Engadget)
• CATL and Octopus Energy Launch 'Swaptopus' to Build European Truck Battery Swap Network (Electrek)
• Spain Orders ISPs to Block Archive.today and Its Mirror Domains (Hacker News)

An AI hallucination nearly triggers a United States military interception of a Chinese vessel, while Apple launches the iPhone 18 Pro lineup worldwide. Plus, Meta expands its Muse desktop agent to Mac, and California pushes for mandated AI kill switches.

Stories covered:

• AI Hallucination of Chinese Nuclear Components Nearly Triggered US Military Attack (Ars Technica)
• Apple Launches iPhone 18 Pro and Apple Watch Series 12 as Foldable 'iPhone Duo' Previews Emerge (9to5Mac)
• Meta's New Muse AI Agent Launches on Mac, Overtaking ChatGPT in App Store (TechCrunch)
• California Governor Gavin Newsom Pushes for Mandated AI 'Kill Switches' (The Verge)
• FCC Allows Gulf Sovereign Wealth Funds to Acquire 49.5% Stake in Paramount-Warner Bros. (Ars Technica)
• FBI and Coast Guard Board Hacked Oil Tankers Heading to US Coast After Systems Compromised (TechCrunch)
• Anthropic Establishes Physical Biology Lab in San Francisco to Conduct Wet-Lab Experiments (TechCrunch)
• Google Bypasses AOSP to Introduce Proprietary APIs in Android 17 (Hacker News)