Welcome back! HoverAir may have found the craziest way possible to dodge the FCC’s drone ban. They’re insisting the new flying camera is not technically a drone but (get this) a pocket camera with optional wings.
The $750 Versa has already been approved by the FCC and launched on Indiegogo, but we’ll see if the loophole lasts under heavier scrutiny. Somewhere out there, a regulatory lawyer is staring at a tiny airborne camera. To be or not to be a drone… that is the question.
Would You Sell Your Phone Data to OpenAI?
If your data is already being used to train AI, why not get paid for it?
That's the premise behind Verb, a new marketplace that lets people sell their personal data directly to AI companies and pocket the cash. Verb says it excludes passwords, text messages, and health data from any analysis.
Here's how it works:
Install Verb's tracker on your phone and choose which categories to share (shopping habits, app usage, social media activity, etc.).
Toggle off anything you don't want collected. Block specific companies from buying your data.
Set your own price—the more you share, the more you can potentially charge.
The earlier version of this product (Verb.AI, aimed at Gen Z) was paying users around $50/month to hand over their scrolling data. The new marketplace opens it to everyone—and importantly, positions AI labs as the biggest buyers.
My POV: I'm honestly not sure whether this is a net good for people. Even with sensitive categories excluded, the privacy implications are real, and the risk of any big personal-data database getting breached is not theoretical. Previous attempts to pay consumers for their own data have struggled to gain traction.
But I do think this is where things are heading.
AI companies need enormous amounts of fresh, real-world data. Publishers like Reddit and Reuters have already realized they can negotiate multi-million dollar licensing deals instead of giving their content away. And people are getting much more sensitive about handing over their own data for nothing in return.
So eventually, the AI companies may just have to start paying us for it too. Verb is early to that shift. It probably won't be alone for long.
Hit reply and let me know: would you install it?
— Matt
OpenAI Adds Privacy-Focused Safety Monitoring
Via Britannica
OpenAI is previewing a new privacy-centric safety system for enterprise customers that can watch for misuse without retaining customer data.
What launched: Private Safety Processing detects potentially harmful use across multiple conversations without storing customer data. OpenAI is currently rolling it out to select customers.
How it works:
Instead of monitoring only one session at a time, Private Safety Processing can look for patterns of abuse across multiple conversations. That matters because a bad actor could try to spread malicious requests over several sessions to avoid detection.
If the system is triggered, OpenAI says it sends a narrowly defined signal about the type of activity, rather than handing over full customer conversations for review.
Anthropic made its move: The play comes shortly after Anthropic announced a 30-day data retention policy for “covered models,” including its Mythos-class systems and future models with similar capabilities.
Safety race: OpenAI’s announcement also lands as the company is publicly slowing some frontier AI development under a voluntary pacing approach. Put together, the message is clear: model quality is now table stakes. They’re competing on who can offer the strongest safety story without scaring away enterprise buyers.
The bigger picture: Safety, privacy, and guardrails are becoming product features. Enterprise customers want powerful models, but they also want proof that their sensitive data will not become training material, investigation material, or someone else’s liability.
Meta AI Gets a Mac App
Meta is launching a dedicated Mac app for its AI chatbot as it pushes Meta AI further into everyday productivity workflows.
What’s new: The Mac app lets users share a window with Meta AI, allowing the chatbot to answer questions, make suggestions, or create content based on what it can see on screen. It also supports dictation across apps.
Productivity push: Meta is positioning the app less like a standalone chatbot and more like a work assistant. The company says Meta AI can now connect with Instagram and Facebook accounts, Meta ad campaigns, and Google Workspace across web, mobile, and Mac.
The bigger picture: The launch brings Meta closer to rivals that already have desktop AI apps. Gemini can share your window, while ChatGPT and Claude go further by letting their apps take control of your computer.
Everyone's arguing about which AI model is smartest…
That's the wrong argument.
An AI can tell an employee what the expense policy says. It can't tell you it's not authorized to approve that request itself, catch the VPN issue causing the lockout, and route it to the right person with a clean audit trail. That gap between reasoning and safely acting is the real story right now.
Think GPS vs. air traffic control. A GPS optimizes one route for one driver. Air traffic control coordinates thousands of aircraft at once, enforcing safety in real time. Most of AI right now is building great GPS. Almost nobody's building air traffic control.
Gartner projects 40% of agentic AI initiatives fail by 2027- not from weak models, but from weak governance. If you're deploying AI right now, this is the piece to read before you scale it: from AI chaos to AI control.
Generate SOPs With AI
Via SOPMaker
SOPMaker generates standard operating procedures for industries including restaurants, manufacturing, healthcare, and HVAC. Users describe a process, and the tool drafts a complete SOP with roles, steps, safety notes, and compliance citations from OSHA, FDA, and ISO standards.
How you can use it
Turn internal processes into structured SOPs
Add roles, steps, safety notes, and compliance references
Review AI edits as tracked revisions
Accept or reject changes before finalizing
Pricing: Free and paid
Optimize for AI and Google Search
Via RankControl
RankControl is an AI-powered SEO platform that helps brands appear in both AI search responses and traditional Google results. It uses seven AI agents to plan, build, and publish content on a brand’s domain, while tracking citations across AI search platforms and Google.
How you can use it
Build content for generative and answer engine optimization
Track AI and Google citations across six search platforms
Improve entity authority and citation patterns
Publish SEO content directly on your domain
Pricing: Paid
Jobs, announcements, and big ideas
OpenAI adds transparent background generation to its GPT-Image-2 API in preview.
Anthropic ships Computer Use, Skills, and Files APIs, bringing Claude agents to production.
Google rolls out a Preferred Sources button letting publishers claw back traffic lost to AI search.
OpenAI launches a feature that allows ChatGPT to read and reply to iMessages.
Pew finds that 35% of webpages published since ChatGPT's launch show signs of AI authorship.
Apple Music will require labels on AI-generated tracks.
Ramp debuts Router, its own AI model routing service, free through the end of 2026.
The best models—and the ones I’ve had the most fun with this week! That and more AI news in my video ⤵️
That’s a wrap! See you next week for more.
