Google rolled out the ultra-cheap frontier model 'Gemini 4 Argon', and the big three labs jointly announced cyber safety cooperation. Also covered: an OpenAI employee's FBI arrest over a listening device, Character.AI's bankruptcy filing, Kakao's KakaoTalk AI assistant reveal, and the plunge in Korean memory stocks.
🚀 Google Launches 'Gemini 4 Argon' — At About One-Fifth of Rivals' Price
Google launched its new frontier model Gemini 4 Argon on October 13. Introductory pricing of $2 per million input tokens and $10 per million output tokens is roughly one-fifth of GPT-6 Astra ($10/$50) and Claude Opus 5.5 ($4/$20), and those prices hold until November 30. It scored well on coding, security, science, finance, and legal benchmarks, and Google also announced a robotics-specialized variant and success on all 12 cyberattack scenarios. Early community reaction has been unusually positive across the board.
🕵️ OpenAI Engineer Arrested by FBI for Planting Listening Device in Meeting Room
OpenAI employee Abdulrahman Alfozan (35, a Saudi national) was arrested by the FBI on October 9, accused of installing an encrypted recording device, hidden in tissue paper, in a USB port behind a TV in an executive conference room at the San Francisco office around September 28. The Department of Justice said he sought to secretly record sensitive internal discussions about 'superintelligence technology', and he appeared in court on October 13. OpenAI claimed it identified the person who planted the device using its own AI models, and fired the employee. If convicted, he faces up to 5 years in prison and a $250,000 fine.
💔 Character.AI Files for Chapter 11 Bankruptcy — Disney to Buy Assets
Character.AI announced it has voluntarily filed for Chapter 11 bankruptcy protection and will pursue an asset sale under court supervision. Disney has agreed to buy some assets for $20 million, and the CEO warned the platform could shut down if no broader bid emerges. The company has effectively been a shell since August 2024, when Google licensed its technology for $2.7 billion and took its founders. A key question in the proceedings is how millions of users' private conversation data will be handled.
🛡️ OpenAI, Anthropic, and Google DeepMind Jointly Announce Cyber AI Safety Effort
OpenAI, Anthropic, and Google DeepMind jointly announced cyber-focused AI models, safety measures, and an access program after weeks of coordination. It is unusual for the three labs — fierce competitors for enterprise customers, research talent, and computing power — to publicly show safety cooperation. Coming alongside the royal-hosted AI safety summit in Scotland, it signals a clear industry trend toward treating security risks as a 'shared response' target.
🇰🇷 Kakao Unveils KakaoTalk AI Assistant 'Kanana Space' — Closed Beta in December
At its developer conference 'if(kakao)26' on October 13, Kakao unveiled 'Kanana Space', an AI assistant for KakaoTalk, and 'Canvas', a PC-based agent creation service. Swiping down KakaoTalk with two fingers summons the Ryan character, which handles sending messages, managing schedules, web searches, and changing settings by voice or text — with plans to expand into shopping, gifts, reservations, and payments. Pre-registration has opened, and a closed beta for some users begins in December. The launch is the centerpiece of Kakao's hyper-personalized AI strategy, pitched as 'understanding connections'.
On October 13, SK Hynix closed at ₩1,443,000, down 10.09%, while Samsung Electronics fell 7.39%. Foreign investors net-sold ₩4.51 trillion, sending the KOSPI down 5.46% to 6,035.35. The sell-off, centered on memory stocks, is deepening as AI memory bubble and supply concerns resurface.
🇰🇷 Samsung's HBM4E Passes Nvidia and Hyperscaler Qualification Tests
Samsung Electronics' next-generation 12-high HBM4E chip has passed customer qualification tests at Nvidia and major global hyperscalers, according to KED Global. The result strengthens Samsung's position in the next-generation AI memory race against SK Hynix and Micron, and the chip is being discussed as a candidate for Nvidia's next-generation accelerators. However, the size of actual supply contracts and mass-production schedules have not yet been confirmed.