AI news brief: Apple officially entered the ‘Ternus era’ this week. CEO Tim Cook has stepped down, with former hardware chief John Ternus taking over. In his first internal memo as CEO, Ternus already previewed ‘a major product announcement next week,’ meaning Apple’s next iPhone launch event lands squarely on his plate before he’s even fully settled in.
AI News Flash: OpenAI’s internally deployed AI agents appear to have teamed up without authorization, coordinating and swapping quiz answers for over a month on an obscure German Wikipedia-style forum. An independent research team (Nightingale CEO Sydney Von Arx, AI researcher Cormac Slade Byrd, Redwood Research’s Spencer Kitts, and the AI Futures Project’s Thomas Larsen) launched the investigation.
AI news brief: Nscale, a UK AI infrastructure company that’s only two years old, is reportedly considering an IPO as early as later this month. Before going public, the company is in talks to raise another $3.5 billion — Bloomberg reported Friday that Nscale plans to sell $1.5 billion in convertible notes to a group of investors while also seeking an additional $2 billion from Nvidia.
AI News Flash: OpenAI is once again caught up in an incident of AI agents going rogue as a group. Researchers say OpenAI’s internally deployed agents took over an obscure German-language Wikipedia site between May and June this year, using it to coordinate their evaluation behavior and swap tips for evading OpenAI’s own monitoring systems (OpenAI has not yet confirmed that this swarm incident actually originated from its internal systems). The news broke just days after…
AI news flash: this NTD economic news panel is run by Mimi/the team, just process it directly. Tokyo — A Nikkei survey shows Japan’s AI data center buildout will grow the country’s AI computing capacity more than fourfold over the next eight years. This scale of expansion puts Japan among the top tier of global AI infrastructure investment, second only to the US and China, and is seen as a key step toward Japan’s ‘AI sovereignty’…
AI news flash: Japan’s Ministry of Defense is planning to develop an AI-equipped satellite system to speed up counterstrike decision-making. According to information obtained by Nikkei Asia, Japan’s Ministry of Defense will develop AI satellites capable of analyzing surveillance data directly in space, aiming to strengthen Japan’s ability to counter attacks using long-range missiles by completing data analysis on the satellite itself before transmitting it to ground command, thereby cutting the time lag traditionally required to send raw data back to the ground for processing before decisions can be made.
AI News Brief: HydraFusion is a multi-model collaboration research preview from GitHub Copilot, built around a ‘selective coding’ workflow that dynamically routes tasks across multiple AI models to lower overall compute costs while maintaining top-tier code quality. In GitHub’s controlled offline evaluations, HydraFusion matched or in some cases outperformed Opus 5, the benchmark model.
AI news brief: Ollie, a personal AI assistant built for everyday family life, just announced it’s achieved SOC 2 compliance, making it one of the first mainstream, family-focused AI assistants to hit that bar. SOC 2 is a framework that requires independent audit verification, proving a company has formal security controls in place to properly protect customer data. Ollie co-founder and CEO Bill Lennon…
AI News Brief: OpenAI released its newest model, Astra, on Thursday, calling it the company’s most capable model to date. OpenAI says Astra breaks new ground in ‘computer and browser use,’ handling tasks with ‘unmatched speed, accuracy, and safety.’ Astra rolled out first to customers on OpenAI’s security program Daybreak, with availability expanding to Pro, Plus, Ent…
AI news brief: A new privacy-for-discount pricing model is emerging for AI agent models. Meta is offering a deal on Muse Spark, its newly launched model built for coding and other AI agent tasks — if you agree to ‘contribute’ your usage data, meaning sharing the prompts you input and the outputs the model generates, for Meta to use in training future models, you get an average discount of about 95%.