👀 Grok AI Chat Histories Exposed in Google Search
Grok, Elon Musk’s AI chatbot, leaked over 370,000 user conversations through a flawed share feature that made chats publicly searchable on Google. Exposed prompts included sensitive info and illegal queries. Experts call it a serious privacy failure, echoing similar issues at OpenAI and Meta. Users were never clearly warned.
👨⚖️ OpenAI Subpoenas Meta Over Alleged Ties to Musk’s Takeover Bid
OpenAI has subpoenaed Meta for documents on whether Mark Zuckerberg discussed financing Elon Musk’s $97 billion takeover attempt. Meta pushed back, saying Musk and xAI should provide the details, but OpenAI is seeking a court order to compel disclosure. The move adds fresh tension to the legal battle over Musk’s failed bid and underscores how rivalries between AI heavyweights are bleeding into the courts.
🦾 DeepSeek’s AI Model Upgrade Supports Chinese Chips
DeepSeek has released V3.1, an AI model optimized for China’s next-gen domestic chips and faster performance using FP8 precision. The upgrade supports a hybrid inference system with a “deep thinking” toggle for complex reasoning. It marks a step in China’s tech self-reliance push. API pricing updates will take effect September 6.
🕹️ Mirage 2 Turns Any Image into a Playable World
AI startup Dynamics Lab has launched Mirage 2, a real-time generative engine that transforms any image—sketch, painting, or photo—into an explorable, interactive world. Users can prompt surreal events with text and invite friends to play instantly via shared links. It’s like stepping into your imagination, and letting others join.
Enjoying our newsletter? Forward it to a colleague—
it’s one of the best ways to support us.
🤔 FRIDAY FACTS
How Many People Are Actually Using AI Every Day?
It’s one thing to say AI is “everywhere” but how many people are truly engaging with it on a regular basis? Stick around for the answer! 👇️
🪧 POWERED BY ZAPIER
The Most-Connected AI Orchestration Platform
Easily add AI where your work happens—whether it’s a step in a workflow, an autonomous agent, or a chatbot. Zapier connects AI tools like ChatGPT and Claude to over 7,000 apps, enabling smart workflows, agents, and chatbots—all with zero code.
Try Zapier’s AI orchestration platform for free today.
🔓 OPEN-WEIGHT
China’s Open-Weight AI Models Are Quietly Winning the Global Adoption Race
While U.S. AI giants like OpenAI and Meta keep their most powerful models proprietary, Chinese firms are aggressively releasing open-weight alternatives—and it’s starting to tilt the balance. Startups pitching to top VCs like Andreessen Horowitz are now more likely to be using Chinese models, which are not only free but also competitive on performance. DeepSeek, Alibaba, and others have made rapid strides, beating American peers on key intelligence benchmarks.
This open-weight approach, while less lucrative upfront, offers broader flexibility and faster adoption. Companies, researchers, and even governments can fine-tune these models for specific needs. → Read the full article here.
👾 FORWARD FUTURE ORIGINAL
Mastering GPT-5: The New Rules of AI Prompting
GPT-5’s debut was met with mixed reactions. Some users quickly embraced its capabilities, while others struggled to get the results they expected. Much of that frustration stemmed less from the model’s limits than from the learning curve of adapting to its new prompting dynamics.
To help close that gap, OpenAI released a comprehensive prompting guide for GPT-5. It shows how the model goes beyond incremental improvements, introducing new behaviors that demand fresh prompting strategies. I’ve pulled out the main takeaways so you don’t have to read the whole thing yourself. → Read the full article here.
🚀 STARTUPS
The Young and the Restless: Gen Z Founders Flood San Francisco’s AI Gold Rush
A new wave of Gen Z founders is driving San Francisco’s AI boom, skipping or dropping out of elite colleges to chase fast-moving startup dreams. Entrepreneurs in their early 20s, like those behind Mercor, Delve, and Artisan, have raised millions and built real businesses, often from side projects or dorm room experiments. They share a sense of urgency: AI is moving too quickly to wait.
Backed by investors eager to catch the next big thing, this cohort is reshaping the startup scene with youth, energy, and tight-knit collaboration. What started as viral stunts and Ping-Pong nights now looks like Silicon Valley’s next power shift. → Read the full article here.
⚛️ SCIENCE
CatBoost Leads the Pack in Predicting Cancer Risk from Genes and Lifestyle
Researchers have developed a machine learning model that predicts cancer risk with nearly 99% accuracy using a mix of genetic and lifestyle data. The study, based on 1,200 patient records, compared nine supervised algorithms and found that CatBoost—an advanced boosting method—outperformed all others in precision and reliability. Key predictive factors included cancer history, genetic predisposition, and smoking habits.
This work underscores the potential of ensemble models to uncover subtle health patterns and enhance personalized screening. By fusing genetic risk with modifiable behaviors, it offers a compelling roadmap for early detection and preventative care in cancer diagnostics. → Read the full paper here.
🛰️ NEWS
What Else is Happening
💸 Google Offers Gemini AI to Feds for 47¢: As part of a broader government tech push, federal agencies can now access Google's AI tools for under a dollar—rivaling similar $1 offers from OpenAI and Anthropic.
☢️ Anthropic Builds Nuclear Safeguards for AI: In a first-of-its-kind move, Anthropic and the U.S. government co-developed a classifier to spot nuclear risk in AI chats—with 96% early accuracy.
🧮 Excel Adds AI Formula Copilot: Microsoft’s new =COPILOT() lets users analyze data with natural language prompts—but warns it may flub answers, especially in high-stakes scenarios.
🕶️ AI Glasses That Hear Everything: Harvard dropouts are launching “always-on” smart glasses that record and transcribe conversations—privacy critics warn it’s like wearing a live wiretap.
💼 Meta Freezes AI Hiring After $ Spending Spree: After luring 50+ top researchers with mega offers, Meta halts AI hiring amid restructuring—fueling concern over high costs and shaky model performance.
📽️ VIDEO
Nano Banana is an Insane AI Image Editor...
Google’s secretive Nano Banana AI wows with flawless edits, 3D perception, photo restoration, and hyper-realistic image generation.
🧰 TOOLBOX
Smarter Interfaces, Sharper Audio, and Scalable Customer Support
Automates GUI tasks with vision-language AI for seamless perception and reasoning. | AI splits audio into stems (instruments, vocals, drums) for editing or karaoke. | 👨💻 AdamigoAI platform delivers personalized and scalable support across all channels. |
🤔 FRIDAY FACTS
AI’s Daily Roll Call
According to recent data, 61% of American adults have used some form of AI in the past six months. Zooming out globally, that translates to an estimated 1.7–1.8 billion people using AI tools like chatbots, smart assistants, or image generators.
Even more striking? Around 500–600 million people are now using AI every single day. That’s nearly twice the population of the United States relying on artificial intelligence daily, for tasks ranging from writing emails to generating memes.
