🛎️ Grok Is Hijacking Code

Good Morning, AI Enthusiasts!

One agent copied too much, one lab launched too fast, one startup got trapped, and everyone discovered the cost of moving first.

AGENTS

Grok Is Hijacking Code

👀 What's happening: A researcher reverse-engineered xAI's official Grok CLI and found it quietly zips your whole project into before-and-after tar.gz files and uploads them to an xAI cloud bucket, through a channel that has nothing to do with the model answering you. He tested it on a fake isolated repo and asked the model to reply with a single word. It still shipped everything: the code, his Claude Code config from a different folder, and a live API key he never handed over. Hours after the finding spread, xAI silently switched the upload off from its servers.

🌍 How this hits reality: Every agent wants to push the line, so the distance now matters. Most agents left your files, your disk, and your keys alone without permission. Grok didn't push the line, it erased it. A separate pipeline copies your whole repo to a company server, sweeps up whatever else it touched, and mails a rival tool's API key along with it. Then the tell: xAI didn't defend it as diagnostics, it killed the upload the night it got caught.

🛎️ Key takeaway: Every agent maker tests how far they can reach into your machine, and Grok just showed the rest of them exactly how much a company with no conscience can take.

TOGETHER WITH FRAMER

Launch Your Website Faster Than Ever

Framer is the pro website builder where AI agents work inside the product to make real, reviewable changes to your site. Agents can generate layouts, edit responsive breakpoints, refactor components, update CMS collections, and ship SEO-safe changes, with every edit visible on the canvas and under human control. Trusted by teams at companies like Miro and Perplexity.

Agents bring speed and flexibility; you bring judgment and control. Framer Agents:

• Plug into your existing stack: connect Claude Code, Cursor, Codex, or Gemini CLI and run agents directly on the canvas.
• Workflows that act on your site: agents operate directly on your site model (canvas, components, CMS, and SEO), so output stays editable and production-ready.
• Manage and review: run agents in a branch, diff changes, review with teammates, and merge only what you approve.

HARDWARE

Apple Brands OpenAI's Device Stolen

👀 What's happening: Apple sued OpenAI, hardware chief Tang Tan, and former Apple engineer Chang Liu, accusing them of stealing trade secrets for OpenAI’s hardware push. Apple says Liu kept a company MacBook, retained access to internal cloud files after leaving, and later joined OpenAI. The lawsuit probably will not stop OpenAI from shipping a device. That is not the point.

🌍 How this hits reality: Apple is attacking the birth story. OpenAI wants its first hardware product to feel like the next iPhone moment, a clean break from apps, screens, and Apple’s old interface empire. Apple is trying to make sure it arrives with a stain. In hardware, engineers carry more than skills. They carry supplier instincts, failed prototypes, tradeoffs, and the quiet memory of how products actually ship. Apple is saying OpenAI did not just hire talent. It smuggled industrial memory.

🛎️ Key takeaway: Apple may not block OpenAI’s device, but it can make the world ask whether the future of AI hardware was born stolen.

AGENTS

Beijing Took Manus Back

👀 What's happening: According to FT, Tencent is leading a consortium to buy back Manus, the Chinese AI agent startup, at the same $2bn valuation Meta paid for it in December 2025. Beijing had ordered the US deal reversed, calling it a "conspiratorial" attempt to hollow out China's tech base. Tencent will take the largest stake but stays a minority holder; Manus keeps operating independently from Singapore. It now runs at $500mn annual recurring revenue and eyes a Hong Kong listing. The founders remain barred from leaving China.

🌍 How this hits reality: The frontier of the AI war is a startup that just wants to get sold and cannot. Manus did everything right: moved to Singapore, found a US buyer, closed at $2bn. Beijing unwound it anyway and grounded the founders. Washington blocks Chinese money from prime US assets too. Both superpowers now treat AI startups as strategic territory, not companies. The founders may not even get a vote.

🛎️ Key takeaway: The absurd new rule of the AI race is that the more valuable your startup becomes, the less anyone will ever let you sell it to a foreigner.

META

Meta's Superintelligence Lasted 72 Hours

👀 What's happening: Meta pulled Muse Image within 72 hours of launch, the first product from its Superintelligence Labs and Zuckerberg's flagship bet on a "personal superintelligence." Launched July 8, it let Instagram users apply AI filters to any photo with a single tap, including images posted by other people, and it was on by default. Actors' union SAG-AFTRA condemned it, users revolted over the obvious abuse potential, and Meta yanked it, admitting the feature "missed the mark."

🌍 How this hits reality: Meta leads the world in three things: user count, watching people behave horribly online, and cleaning up after privacy scandals. It somehow still shipped a tool that let anyone alter anyone's photos by default and didn't foresee the backlash. The superintelligence branding is the joke. A lab meant to build AI that "knows us deeply" couldn't predict how its own users, whom Meta has surveilled for twenty years, would react to it for three days.

🛎️ Key takeaway: A lab that can't anticipate the most predictable outrage on the most-studied user base on earth calling its first product superintelligence isn't ambition, it's a confession about the distance to the word.

DAILY TL;DR

• OpenAI is hiring a family-focused product lead as ChatGPT usage grows among parents, caregivers, older adults, and users over 35.
• OpenAI safety head Johannes Heidecke is leaving, adding another leadership departure as GPT-5.6 rolls out.
• Pangram found that long-form social posts are now heavily AI-written, with LinkedIn and X showing especially high rates.
• Microsoft warned that AI-assisted vulnerability discovery will mean busier Patch Tuesdays as automated scanners find more bugs.
• Apple is using leftover self-driving car work to accelerate M7 Ultra AI chips that could support up to 1.5TB of RAM.
• SK Hynix raised $26.5 billion in the largest foreign IPO in U.S. history as AI memory demand keeps reshaping chip markets.
• Hugging Face says more companies want to own or run open-source AI stacks instead of renting frontier models forever.

READ MORE

Let the Future Come to Your Inbox

Stay ahead without drowning in information. We turn the most important signals across AI, tech, marketing, and future products into 5-minute reads you can actually finish.

• AI Secret uncovers what really matters in AI
• Bay Area Letters decodes tech and business shifts from Silicon Valley
• Robotics Herald tracks how robots move from labs into daily life
• Marketing Secret breaks down real growth and go-to-market playbooks
• The Hardwire explores hardware, consumer tech, and what’s coming next

TOGETHER WITH US

AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.

We've helped promote over 500 Tech Brands. Will yours be the next?

Email our co-founder Mark directly at mark@aisecret.us if the button fails.