xAI co-founder exits to launch AI VC firm

The tech landscape shifts rapidly as a key xAI executive departs to launch his own venture capital firm, the US government implements covert tracking measures on AI chip shipments, and OpenAI reverses course on its controversial model selection changes. Today’s digest also covers quantum computing breakthroughs, Apple-Musk tensions, and Google’s massive infrastructure investment.

xAI Co-founder Igor Babuschkin Launches AI-Focused VC Firm

Igor Babuschkin, who co-founded Elon Musk’s xAI in 2023 and led its engineering teams, is departing to establish Babuschkin Ventures, a firm dedicated to backing AI safety research and startups with a mission to advance humanity. His exit follows months of controversy at xAI over Grok’s biased and offensive outputs, though the company’s AI models remain competitive with OpenAI, Anthropic, and Google DeepMind. Babuschkin, who previously worked at Google DeepMind on AlphaStar and at OpenAI, credits Musk with teaching him to personally tackle technical problems and work with urgency.

US Secretly Embeds Trackers in China-Bound AI Chips

The U.S. government is secretly inserting location trackers into select shipments of advanced AI chips to catch smugglers before hardware is illegally rerouted to destinations like China. These trackers have been found hidden in packaging or directly inside servers from Dell and Super Micro, containing targeted AI hardware produced by both Nvidia and AMD. Aware of the risk, some China-based resellers now routinely inspect diverted shipments for hidden devices, with one smuggler warning another in a message to “look for it carefully.”

OpenAI Restores GPT-4o as Default Model After User Backlash

Following significant user backlash to its deprecation last week, OpenAI has restored GPT-4o as the default choice in the model picker for all paid ChatGPT subscribers. The company also introduced new “Auto”, “Fast”, and “Thinking” settings for GPT-5, giving people direct options to bypass the model router that was meant to simplify the user experience. Sam Altman acknowledged the rough rollout, promising more customization for model personality and giving plenty of advance notice before the company considers deprecating GPT-4o in the future.

IBM and Google Claim Quantum Computing Breakthrough Is Near

IBM published its quantum computer blueprint and now claims it has “cracked the code” to build full-scale machines, with the company’s quantum head believing they can deliver a device by 2030. While Google demonstrated error correction using surface code technology that needs a million qubits, IBM pivoted to low-density parity-check codes which it says require 90 percent fewer qubits. The competition is expanding as IonQ raised $1 billion to target 2 million physical qubits by 2030, while Nvidia’s CEO sparked investor rallies in other quantum computing stocks.

Apple Denies Musk’s Claims of App Store Bias Toward OpenAI

Apple rejected Musk’s accusation that the App Store makes it impossible for AI companies besides OpenAI to reach the top spot, citing objective criteria for featuring thousands of apps. Musk’s threat to sue comes after Apple’s partnership with OpenAI to integrate ChatGPT into Siri and iOS, though AI apps like Perplexity and DeepSeek have also topped charts. The dispute comes amid heightened regulatory scrutiny of Apple’s app distribution practices, including ongoing legal pressure from the Epic Games case.

AirPods to Get Live Translation with iOS 26

The newest iOS 26 developer beta 6 includes references for a real-time translation function, letting AirPods wearers hear speech from another tongue converted directly into their own native language. An image inside the beta suggests English, French, German, and Portuguese as the first choices, with support confirmed for the upcoming AirPods Pro 2 and AirPods 4 models. This capability would expand on similar tools in Apple’s Phone, Messages, and FaceTime apps and could become mainstream due to the company’s large wireless headphone market share.

DeepSeek Delays R2 AI Model Due to Huawei Chip Setbacks

Chinese AI startup DeepSeek postponed its R2 model launch after technical issues arose when training with Huawei’s Ascend chips, switching to Nvidia chips for training and Ascend for inference. The delay underscores challenges for Chinese AI firms trying to reduce reliance on U.S. technology amid chip export restrictions and security concerns. R2 is now expected in the coming weeks, following the R1 model’s success and competition from major players like ByteDance, Tencent, and Alibaba.

Google Invests $9 Billion in Oklahoma AI and Cloud Expansion

Google will spend $9 billion over two years to build a new data center in Stillwater and expand its Pryor facility, boosting U.S. AI and cloud capacity. The investment includes workforce and education initiatives, following a $1 billion commitment to AI training for U.S. universities and nonprofits. The move aligns with Big Tech’s race to scale AI infrastructure amid rising competition and U.S. onshoring efforts.

Today’s developments highlight the ongoing turbulence in the AI sector, from executive departures and government surveillance to infrastructure investments and technological breakthroughs. As companies navigate chip restrictions, user demands, and competitive pressures, the race for AI dominance continues to reshape the tech landscape.