Welcome, AI Pioneers!
If you use Copilot at work, read this. A simple prompt let insiders pull summaries of confidential files without showing up in the logs. Meanwhile workers want AI as a teammate, not a replacement. Nvidia is building a China-compliant GPU. OpenAI is planting roots in India.
📌 In today’s Generative AI Newsletter:
Copilot audit gap let activity go unlogged
Stanford workers prefer partnership over automation
Nvidia preps B30A for China
OpenAI opens New Delhi office
Special highlight from our network
AI is already shaping how boards govern, allocate capital, and manage risk.
Join mindit.io’s live webinar Demystifying Modern AI for Board Efficiency on August 27, 2 PM CEST, and learn how AI is reshaping the boardroom agenda.
In 45 minutes, you’ll gain clarity on:
Where AI is driving measurable board-level impact
Case studies of adoption in strategic leadership
Risks and blind spots that boards must address
How to become an AI-literate leader without technical training
If you’re a board member, advisor, or aspiring leader, this session will help you scale your impact.
🕵️ Microsoft Copilot's Logging Flaw Exposed
Credit: cybersecuritynews
A critical hole in Microsoft Copilot allowed users to quietly bypass access logs, raising alarms about the integrity of enterprise AI security. The flaw, uncovered by Zack Korman, CTO at Pistachio, meant insiders could ask Copilot to summarize confidential files and leave no trace in the audit logs.
Here’s what went wrong:
Ghost activity: By instructing Copilot not to reference the file it summarized, the log’s “AccessedResources” field was left blank. No record, no trail.
Exploitable vector: A malicious insider could have siphoned sensitive documents undetected, undermining the very compliance standards Microsoft markets Copilot on.
Pattern of weakness: Researchers have repeatedly flagged prompt injection vulnerabilities in Copilot, including takeover scenarios demonstrated at Black Hat 2024.
Silent patch: Microsoft fixed the flaw on Aug. 17, 2025, but declined to assign a CVE, frustrating researchers who described the disclosure process as opaque.
The bug serves as a reminder that AI tools can become blind spots in compliance systems, which is more than just a minor inconvenience for businesses integrating Copilot into their everyday operations. Trust in enterprise AI won’t hinge on flashy integrations, but on whether its guardrails can withstand deliberate manipulation.
📊 Stanford Finds We’re Automating the Wrong Jobs
A new Stanford survey of 1,500 workers and AI experts reveals a sharp mismatch between where AI is being built and what workers actually want. Instead of full replacement, people overwhelmingly prefer partnership with AI and the skills rising in value aren’t the ones Silicon Valley predicted.
Here’s what the study uncovered:
Workers want teammates, not replacements: 45.2% of occupations prefer “H3 equal partnership” with AI, while only 17% of creative tasks scored positively for automation.
Investments are misaligned: 41% of startup funding is flowing into zones workers don’t want automated or that have low technical feasibility.
The wage hierarchy is flipping: High-paying “analyzing data” tasks are losing value, while interpersonal communication and training others are becoming premium.
Freedom, not layoffs: Nearly 70% of workers want AI to automate busywork so they can focus on high-value tasks, reduce stress, and preserve creative control.
The AI “takeover” isn’t happening where the hype says it will. Workers want AI to sweep the floors, not paint the canvas. The companies that listen will own the future of human-AI collaboration.
🧩Nvidia Builds New China-Only AI Chip Amid Washington Uncertainty
Credit: Andy Wong/AP
Nvidia is quietly developing a new AI chip for China, codenamed B30A, as the Biden administration weighs whether to loosen restrictions on advanced US chips reaching Beijing. The processor is based on Nvidia’s latest Blackwell architecture and is expected to outperform the current China-approved H20, even as lawmakers warn that any upgrade could tilt the AI race.
Here’s what we know:
B30A specs: Designed with a single-die layout, high-bandwidth memory, and NVLink interconnects, the chip will be less powerful than the flagship B300 but still capable of handling high-end AI workloads.
Release timeline: Early samples may reach Chinese customers as soon as next month, though Nvidia has not finalized full specifications.
Market stakes: China accounted for 13% of Nvidia’s revenue last year. Losing that market could push buyers toward Huawei, whose chips are improving but still lag behind in software support and memory performance.
Security tension: Chinese regulators have warned local firms about security risks tied to Nvidia’s H20, while Nvidia insists its products are built for commercial use with government approval.
Alongside the B30A, Nvidia is preparing the RTX6000D, a weaker inference-focused chip tailored to stay under US export thresholds. Together, the moves show how Nvidia is walking a tightrope: trying to hold
🏗️ OpenAI Plants Its Flag in India with First New Delhi Office
Photo: SVEN HOPPE
OpenAI is cementing its presence in India with a decisive move: its first office in New Delhi. The announcement comes just days after the launch of ChatGPT Go, a ₹399 plan designed for Indian users, marking the company’s most aggressive push yet into a market of nearly a billion internet users.
Here’s what’s unfolding:
Local hiring: OpenAI has registered as a legal entity in India and begun building a New Delhi team to work with policymakers, businesses, developers, and academics.
Leadership hires: The company previously brought in Pragya Misra, ex-Meta and Truecaller, as its public policy and partnerships lead, and former Twitter India head Rishi Jaitly as a senior advisor.
Market strategy: India is ChatGPT’s second-largest user base, where weekly active users have quadrupled over the past year, led by students. The new Go plan undercuts rivals with sub-$5 pricing integrated into UPI.
Upcoming events: OpenAI will host its first Education Summit in India this month and its first Developer Day later this year, signaling deeper local engagement.
India presents both scale and complexity. The country has the world’s largest student base for ChatGPT, a strong government AI mission, and a massive digital payments system. But OpenAI also faces lawsuits from publishers over alleged copyright use, and competition from Google’s Gemini and Perplexity, which have secured telco partnerships and free premium access for millions.
Invest in GenAI Works by Aug 28 and Earn Up to 25% Bonus Shares
GenAI Works is building the infrastructure for the AI agent economy. Invest Now
The AI agent market could reach $52 billion by 2030*, but most people don’t know which startups will survive. Instead of betting on one, you can invest in the orchestration layer every agent will need. Invest by August 28 to earn up to 25% bonus shares
We have built the world’s largest AI community with 13 million members and 2,500+ tools in use. Now we are turning that community into the operating system for AI agents:
The GenAI Protocol: an open source engine for agents to collaborate across tools and workflows.
The Distribution Engine: our first application, turning a prompt into a full workflow from finding customers to launching campaigns.
This is not just another AI app. It is the infrastructure powering them all. Invest by August 28 to earn up to 25% bonus shares
Only one week left. Invest by August 28 to earn up to 25% bonus shares. Invest Now
Reg CF offering via DealMaker Securities. Investing is risky.
*MarketsandMarkets, “AI agents market worth $52.62 billion by 2030,” PR Newswire, July 10, 2024.
🚀 Boost your business with us. Advertise where 12M+ AI leaders engage!
🌟 Sign up for the first AI Hub in the world.
