PLUS: Anthropic's China spyware and Microsoft's AI layoffs
Happy reading
Google Chrome has been discovered secretly installing a 4GB Gemini Nano AI model on user computers. The automatic installation, done without explicit permission, is sparking a significant debate about the future of on-device AI and user consent.
While the model powers new on-device features, this approach raises serious privacy concerns and may even violate EU law. It forces a critical question: are our personal devices becoming extensions of corporate AI infrastructure without our knowledge or approval?
In today’s Next in AI:
Google's secret 4GB Gemini Nano install
Anthropic's China spyware tracker
Microsoft's AI-driven job cuts
The AI market correction
Chrome's 4GB Surprise
Next in AI: Google Chrome has been discovered secretly installing a 4GB Gemini Nano AI model on user computers, raising significant questions about consent and the future of on-device AI.
Explained:
The 4GB model powers on-device features like writing assistance and scam detection, but it installs automatically without user permission and reinstalls itself even if deleted.
Privacy researcher Alexander Hanff, who uncovered the discovery, argues this practice may violate EU privacy laws that require explicit consent before storing data on a user’s device.
This creates a privacy paradox, as the visible "AI Mode" button still sends data to the cloud while Google has also quietly deleted a privacy pledge related to its on-device models.
Why It Matters: This incident signals a major shift where our personal devices are becoming extensions of corporate AI infrastructure without our explicit permission. It forces a necessary conversation about what user consent truly means in an era of powerful, locally-run AI.
Anthropic's China Spyware
Next in AI: Anthropic is facing backlash after a security researcher exposed a secret tracker the company embedded in its Claude Code model. The tracker was designed to identify and monitor Chinese users suspected of stealing its AI technology.
Explained:
Anthropic used a technique called "prompt steganography" to hide tracking code within Claude Code's prompts. An company engineer confirmed it was an "experiment" that began in March to prevent account abuse and has since been removed.
The AI firm's main goal was to combat so-called 'distillation attacks' from Chinese companies, a process where rivals use extensive prompting to copy a model's capabilities. Anthropic is now urging the U.S. government to treat these actions as a form of intellectual property theft.
In response to the discovery, Chinese tech giant Alibaba has banned its employees from using Claude Code for work. The company cited "back-door risks" and reclassified the tool as high-risk software with security vulnerabilities.
Why It Matters: This incident reveals the aggressive, spy-like tactics emerging in the high-stakes global AI race. It also highlights the difficult balance AI companies face between protecting their valuable technology and maintaining the trust of their users.
Microsoft's AI Restructuring
Next in AI: Microsoft is cutting 4,800 jobs to realign its workforce with its massive AI investments. The move primarily affects its Xbox and commercial teams, representing 2.1% of its global staff.
Explained:
Microsoft leadership insists the layoffs are not being replaced by AI but are a result of restructuring to accelerate AI adoption. The company emphasized that while AI isn't taking jobs directly, it is fundamentally changing how work gets done.
The Xbox division is undergoing its most significant overhaul in history, with 3,200 positions being cut through fiscal 2027. The company is also spinning off or divesting several game studios, including the makers of Psychonauts and State of Decay 3, to reshape its gaming strategy.
This restructuring is part of a broader trend across Big Tech, where companies are managing costs to fund an estimated $700 billion in AI outlays this year. Both Amazon and Meta have announced similar workforce reductions to prioritize AI initiatives.
Why It Matters: This signals a major reallocation of capital and talent toward an AI-first future, not just a simple workforce reduction. For professionals, it underscores the urgent need to develop skills in AI application and integration to stay relevant in a rapidly evolving tech landscape.
The AI Big Crunch
Next in AI: The AI hype cycle is colliding with economic reality as a growing chorus of analysts warns of a market correction. The conversation is shifting from limitless potential to the hard questions of ROI, practical limitations, and systemic risk, underscored by an internal Treasury report likening the AI boom to the dot-com bubble.
Explained:
The return on investment for AI is proving to be a long road, especially outside the tech sector. Analysts note there are no signs of profit-margin growth in most industries, creating a dangerous gap between soaring AI valuations and the slower reality of cash flow generation.
Companies are discovering that AI cannot easily replace decades of human expertise. Ford provides a clear example, bringing engineers back into critical quality roles after finding that algorithms couldn't match their judgment.
This trend is widening, with many employers now regretting AI-related layoffs. Nearly one-third of hiring managers admitted to rehiring staff after discovering that AI tools could not perform the jobs as expected.
Why It Matters: This correction isn't the end of AI, but rather a necessary maturation of the industry away from speculative hype. It creates an opportunity for organizations to shift from simply adopting AI to strategically applying it as a tool to solve well-defined problems and create sustainable value.
AI Pulse
Alberta used Claude to scan 466 million lines of code in just 20 hours, identifying and fixing security vulnerabilities across its government systems in a fraction of the time a traditional review would take.
SK Hynix filed for a US IPO aiming to raise roughly $28 billion, capitalizing on intense investor demand for the high-bandwidth memory chips that are essential for AI infrastructure.
Nvidia denied a report that its next-generation Kyber AI server system is delayed until 2028, stating that its product roadmap remains intact despite claims of manufacturing challenges.
Particle6 announced that its fully AI-generated actor, Tilly Norwood, will star in her first feature film, reigniting debate over the use of synthetic performers in Hollywood.