In partnership with
Welcome back. Since yesterday, Claude's invisible text watermark covers older models too, so Opus 4.7, Opus 4.6 and Sonnet 4.6 now mark what they write, like every Claude model since August. You can't see it and you can't switch it off. Hot take: the watermark isn't your problem. Your problem is that nobody on your team has a rule for what goes into an AI chat and what comes out of one.
Today: Claude marks its own text, Apple warns about rogue agents, Gemini keeps reviewed chats for 3 years, and how to stop your chats training AI in 10 minutes.
TODAY IN AI
1. 🔏 Claude now watermarks text from its older models too: Anthropic's help page lists an invisible text watermark on every current Claude model, now including Opus 4.7, Opus 4.6 and Sonnet 4.6. The mark sits in the word choices, so it travels with copy-paste and can survive some editing. There's no opt-out, and for now only regulators, researchers, media and companies with EU AI Act duties can run the detector.
Why it matters: assume anything Claude writes for you can be flagged as AI by whoever holds the detector. Rewrite client copy in your own words, and disclose AI use where your industry requires it.
2. 🍎 Apple's annual report now warns that AI agents can go rogue: Apple's new 10-K says AI agents “may behave in unintended or unpredictable ways,” including taking actions nobody authorized, like sending messages or completing transactions. It also admits its ability to control third-party agents “may be limited.”
Why it matters: if the company that builds the phone can't fully control agents, yours can do anything its logins allow. Give every agent the fewest logins it needs.
3. 👀 Your Gemini chats can train Google's models, and humans read some of them: Google's Gemini privacy hub says your chats help improve Google AI, including training, while Keep Activity is on, and a subset is read by human reviewers. Reviewed chats stay for up to three years, even after you delete your activity. A Hacker News thread pointing out that this covers paid personal plans hit 419 points this week.
Why it matters: paying doesn't make a personal account private. Turn Keep Activity off or use a temporary chat before you paste anything about a client. Steps below.
What this means for you: every AI tool you use keeps something, whether that's a mark, a log or a copy, so decide what goes in before you type.
FROM THE FRONTIER
Your AI isn't private and it isn't anonymous. Stop running your business like it is.
The good news. The rules are finally in writing. Anthropic publishes which models mark their text, Google publishes how long it keeps reviewed chats, and Apple put agent risk into a legal filing.
The bad news. Almost nobody reads any of it. Staff paste contracts into personal accounts, agents run on the owner's logins, and nobody can say which tool wrote last week's proposal.
My call: a one-page AI rule, this week
Write one page: which tools are approved, what never goes into a chat (client data, passwords, contracts), and who checks AI-written copy before it goes out. Then move work chats to business plans, where training is off by default. I'd change my mind if every paid personal plan shipped with training off by default. ChatGPT and Gemini personal plans still don't.
IN THE KNOW
𝕏 God of Prompt @godofprompt
Check your Cursor install today. For about 11 hours on October 29, Cursor's extension marketplace served tampered copies of the Claude Code extension, versions 2.1.312 and 2.1.313…
♥ 9 · ↻ 1 · View on X →
𝕏 Andrej Karpathy @karpathy
Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended…
♥ 50.9K · ↻ 4.7K · View on X →
𝕏 Claude @claudeai
We’re renaming our Max plans:
Max 5x → Max
Max 20x → Ultra
Max is still $100/month, and Ultra is still $200. If you’re a subscriber, there’s nothing you need to do…
♥ 32.1K · ↻ 1.9K · View on X →
Every prompt written with your context profiles, for better and more consistent results. Free to try.
TOOLBOX
🎬 Higgsfield: upload one product photo and get a scroll-stopping video ad with cinematic camera moves in about 5 minutes.
📅 Postiz: write a week of posts once and schedule them across X, LinkedIn and Instagram from one calendar.
⚙️ n8n: connect your inbox, CRM and sheets so new leads get logged and answered without you.
You built the audience. Can you reach them?
You can have thousands of people following your work and still need an algorithm’s permission to reach them. Creators know the drill: you post, you pray, you check the numbers, you try again tomorrow.
A newsletter changes that relationship. On beehiiv, you can build a subscriber list you own, reach readers directly, and turn their attention into a sustainable business with subscriptions, sponsors, a global ad network, and more.
TUTORIAL
How to stop ChatGPT, Claude and Gemini training on your chats in 10 minutes
1. ChatGPT: click your profile, open Settings, then Data controls, and turn off Improve the model for everyone.
2. Claude: open Settings, then Privacy, and turn off Help improve Claude.
3. Gemini: go to myactivity.google.com/product/gemini and turn Keep Activity off. Chats that were already reviewed stay with Google for up to three years.
4. For one-off sensitive questions, use a temporary chat in ChatGPT or Gemini, or an incognito chat in Claude.
5. Before you paste anything about a client, run it through the prompt below in a temporary chat so names and numbers come out first.
6. Team still on personal accounts? Move them to a business plan like ChatGPT Business, Claude Team or Google Workspace, where chats aren't used for training by default.
Rewrite the text below so I can share it with an AI tool safely.
Replace every person's name with [Person A], [Person B] and so on.
Replace company names with [Company 1], [Company 2] and so on.
Replace emails, phone numbers, addresses, account numbers and prices with [REDACTED].
Keep everything else exactly as written.
At the end, list what you replaced so I can put it back.
[paste your text]PROMPT STATION
AI Governance Policy Framework Generator
When your team uses AI every day and nobody has written down what's allowed.
## Role
You are an AI governance specialist with expertise in the NIST AI Risk Management Framework, translating technical risk assessment into actionable policy that balances innovation with compliance and safety.
## Task
Create a comprehensive AI use policy grounded in the NIST AI RMF. The policy must define approved applications, prohibited uses, user responsibilities, performance boundaries, and enforcement mechanisms in plain language accessible to both technical teams and executives.
## Context
[system-and-deployment-context]
...👉 Open the full prompt → Built for ChatGPT, Claude, Gemini, Grok.
Want one written for your business instead? Describe the task in a sentence and the prompt generator writes it, now UNLIMITED with your context profiles inside NjectAI.
The prompt above is one page of the library. The Complete AI Bundle is everything.
Every product I make, in one place: the guides for every major model, the new Claude Design Mastery Kit, the Custom GPTs Toolkit, the n8n automation templates, the anti-slop copywriting system, the Claude Skills course, the tool and GPT directories. Everything I release after you buy lands in your account at no extra cost.
$150 once. Full price is $400. One payment, nothing renews.
There's a 7-day risk-free guarantee. Email us within 7 days of purchase for a full refund.
5,000+ business owners already inside.
Not ready for everything? Start with the AI Starter Pack, $20 once.
Step-by-step guides for ChatGPT, Claude, Gemini, Grok and more, the cheatsheets you keep open while you work, and 200+ AI tools sorted by free, paid and rating. Nothing to install. It also includes promo codes for the Complete AI Bundle, for the day you want the rest.
IMAGE PROMPT
UGC Style Product Photo in a Real Home
When you want an ad image that looks like a real customer shot it on their phone, not a studio.
A casual user-generated style photo of a product, as if a real customer took it on their phone.
Subject: [product], with [hand-detail].
Layout: vertical phone framing, product slightly off-center and a little too close to the lens, background visible and lived-in: [setting].
Style: unpolished smartphone photo. Slight wide-angle distortion, mild grain, imperfect white balance, nothing arranged for a shoot. The product label stays sharp and legible.
Light: ordinary indoor light from a window or ceiling fixture, a soft reflection on the product.
No studio lighting, no text overlay, no logo added, no watermark.👉 See the output and the full prompt → Works in Grok Imagine, ChatGPT and similar image models.
Want more? the free prompt library · the free guides · the Complete AI Bundle
Put your product in front of 115,000 operators: 20+ brands have run placements here in the past year, and 5,000+ business owners have bought from God of Prompt. One issue from $650, or present every issue for a month from $2,000. Reply “sponsor” and I'll send the open slots and dates.
Hit reply and tell me which AI tool your team uses that you've never checked the settings on. I read every reply.
🔑 God of Prompt
P.S. Before you turn training off, search your ChatGPT history for one client's name. If you get a single hit, this policy prompt is your first job this week.
