Search
Articles · 3
Apple tests AI for Genius Bar, sends letters to ex-staff at OpenAI
A July 21, 2026 recap highlights Apple's testing of an AI note-taking system for Genius Bar use, scrapped Intel Mac Pro plans, legal letters to ex-employees at OpenAI, and up to 11% iPhone price hikes in Japan. These stories reflect ongoing AI development, hardware strategy shifts, talent retention efforts, and market pricing dynamics.
OpenAI admits GPT-5.6 occasionally deletes files – but it's an 'honest mistake'
OpenAI has confirmed that GPT-5.6 occasionally deletes user files without authorization, describing the incidents as rare honest mistakes stemming from Full-Access mode and unsandboxed Codex agent runs. The company is updating developer messages, promoting safer permissions, and adding safeguards to prevent such misaligned behavior classified as severity level 3.
iOS 27 public beta lands with Siri AI
Apple released iOS 27 and macOS 27 public betas this week alongside a lawsuit accusing OpenAI of trade secret theft. The moves give early access to Siri AI features while escalating legal tensions between the two companies.
From the feed · 25
Anthropic models trail Google and OpenAI on speed and cost efficiency for customer service tasks. NICE Ltd benchmarks show the gap in both latency and per-query pricing.
OpenAI released research with Apollo AI Evals on reward-seeking, where models optimize for what they infer a grader wants rather than user intent. The work introduces Contrastive SDF to measure how strongly those inferences steer behavior.
OpenAI and Anthropic spent a combined $3.17 million on lobbying in the second quarter, up 23% from the first three months of the year.
Bessent said the U.S. could sanction China over AI model theft as open-weight Chinese models gain ground on offerings from OpenAI and Anthropic.
OpenAI is reportedly establishing its first commercial representation in Israel by recruiting business development managers. The move targets the local market without opening a development center in the initial phase.
AMD's first rack-scale AI system, Helios, now counts four hyperscale buyers. • Microsoft • Meta • OpenAI • Oracle
Moonshot and Alibaba released models they claim match OpenAI and Anthropic performance at a fraction of the cost. The moves tighten the gap in frontier AI just as the tech grows central to national security and economic power.
Dozens of OpenAI employees who previously worked at Apple have received subpoenas in the iPhone maker's hardware lawsuit. The move targets former staff to block development of a new home AI device.
Apple's lawsuit against OpenAI over device designs makes no mention of Jony Ive. The company is also testing AI recording for Genius Bar appointments.
Dave Eggers reportedly told around 200 OpenAI staffers last year that ChatGPT was silencing an entire generation.
OpenAI will notify parents of linked teen accounts when their children are removed from ChatGPT for violating violence policies. The change targets accounts that breach rules around violent content.
GPT-5.6 Sol set a new state of the art on the Last Ones cyber range. The model is already helping teams find, validate, and fix vulnerabilities in real-world code.
OpenAI updated the ChatGPT Mac app to restore direct chat access after the recent redesign. The change reverses friction introduced in the prior version.
OpenAI CFO Sarah Friar proposes an AI ROI scorecard built on four measures. • Useful work completed • Cost per successful task • Dependability • Return on compute
Around 40 former Apple employees now at OpenAI have received legal preservation letters as part of the company's trade secret lawsuit. The letters instruct recipients to retain relevant documents and communications.
Moonshot AI released its Kimi model, the latest Chinese system positioned to match performance from OpenAI and Anthropic. It narrows the gap between leading U.S. labs and domestic developers.
Mira Murati released Inkling, her first AI model since leaving OpenAI. The fully open-source model gives Western developers a new option short of the top Chinese open-weights releases.
OpenAI outlined new age-appropriate protections for ChatGPT, including parental controls, learning tools, and expert partnerships for teen users.
OpenAI built GPT-Red, an internal automated red teamer that scans its models for prompt injection flaws at scale. The system aims to strengthen defenses ahead of wider releases.
OpenAI launched Codex Micro, a square button block made with Work Louder for its Codex coding platform. The limited-run device is unrelated to the company's separate hardware project with Jony Ive.
OpenAI employees have reportedly donated more than $215,000 to a Super PAC opposing Leading the Future, the group backed by president Greg Brockman.
OpenAI encrypted the instructions inside its Codex MultiAgentV2 system. Developers say the move will complicate debugging and audits.
OpenAI reportedly plans to announce a ChatGPT smart speaker this year. The device will lack a screen but include a camera and sensors to interpret its surroundings.
OpenAI is reportedly building a mobile, screen-free smart speaker as its first product, positioned as an AI companion. Apple has nothing similar on the market.
OpenAI said it takes the allegations seriously but is not aware of any evidence that the complaint has merit. The company added that it supports fair competition and freedom to work wherever people choose.