Welcome, humans.

University of Kentucky employee David Durman secured a religious accommodation to avoid personally using AI at work. His belief system? Humanism, which centers human reason and wellbeing without requiring belief in a god. Durman argued that outsourcing his thinking conflicted with his moral responsibility.

Now, other employees with similar beliefs have an example to cite when requesting their own exemption. Just remember, his employer said yes. A court hasn’t given everyone a pass. The exemption covers his own AI use, but he still has to cooperate with colleagues who use it, including in meetings.

An exemption from AI, but not from meetings? HR does have limits.

Here’s what happened in AI today:

Your AI is reading ads behind your back

Congratulations! Your AI assistant is officially part of a coveted demographic: robots with purchasing influence.

That’s right: Publishers are serving ads to AI agents. TIME is already doing it through a company called Mobian, which delivers machine-readable pages containing labeled sponsored blocks to registered agents.

The ads are labeled for the agent. They’re invisible to you because you aren’t reading that version of the page. A brand is paying to get its message into the material your assistant encounters before it gives you a recommendation.

Publishers want to get paid even when you never visit. In a sample, European publishers received 227 AI scrapes for every human visitor referred by AI in the second quarter of 2026. That was up from 150 scrapes per referral the quarter before.

An assistant can read the article and give you the answer without sending you to the website, where the publisher typically earns money from ads or subscriptions. Agent ads offer a replacement route: charge a brand to put its message in front of the assistant doing the reading.

And advertisers are finding other ways into the conversation:

Google announced ads in AI Mode follow-up queries in the U.S. this week, alongside expansion to 10 additional countries.
More than 100,000 advertisers have run ChatGPT ads. OpenAI’s ad system follows the conversation as your needs become clearer.
OpenAI is testing Sponsored Agents: click an ad and start a separate, labeled chat with a business. Now the sales pitch answers questions.

These formats put advertising closer to the moment you’re deciding what to buy. Google says its placements are labeled. OpenAI says its ad-selection system runs separately and doesn’t influence ChatGPT’s answers.

Anthropic has taken the opposite position: Claude will remain ad-free, with the business funded through enterprise contracts and paid subscriptions. So ads are a business choice, not an unavoidable part of using AI.

But an ad-free assistant can still encounter sponsored material on the web. That’s what makes the publisher approach different: the advertising money goes to the publisher, while the assistant reads the paid placement on your behalf.

And what happens when agents start passing that information to each other? Imagine one summarizes a sponsored product claim but drops the label. Another uses that summary to choose what to buy. A sales pitch could start looking like independent research. If the original claim is wrong, that mistake could carry over.

Kevin Browne described the broader problem last year as “verification debt”: unchecked AI output gets reused, and the work of finding and fixing mistakes piles up. Three agents repeating the same claim isn’t three sources.

Why it matters: A sponsored message isn’t automatically false. But the more an assistant does without us watching, the more we need to know where its information came from, who paid to put it there, and what it actually checked before acting. Otherwise, we’re delegating the decision and keeping the bill.

Even the machines can't escape the sponsored content.

🎓 AI Skill of the Day: Turn a screen recording into a handoff guide

You know how to pull the weekly report. Your colleague doesn’t. Somehow, writing down the steps feels like more work than just doing it yourself again. Congratulations, you’re now a recurring calendar dependency.

Here’s a workflow to try: record yourself doing the task, then ask Gemini to turn the recording into instructions. Gemini supports uploaded video analysis, so it can use what’s visible on screen alongside your explanation. Narrating why you choose a particular filter gives it context the clicks alone wouldn’t convey.

Record one short task, such as exporting last month’s sales report, and explain your choices aloud; pause briefly on menus and settings.
Keep the clip under five minutes, then upload it through Add files in Gemini and paste the prompt below.
Follow the resulting guide once without watching the recording, and correct anything missing before sharing it.

Turn this screen recording into a step-by-step guide for a colleague doing this task for the first time.

Include the goal, anything they need before starting, and numbered steps using the exact button and field names visible on screen. Add timestamps so I can check each step.

Explain choices I narrated. If a step or label is unclear, mark it “needs checking” rather than guessing. Finish with how to confirm the task is complete.

Pay particular attention to anything you did automatically, like selecting the right account or changing a date range. That’s often the step someone else needs most.

Have a specific skill you want to learn? Request it here.

🍪 Treats to Try

Google SynthID Detector checks your images, videos and audio for invisible AI watermarks from participating companies, including OpenAI, though it won’t catch every AI creation.
Gemini’s File Editor creates and edits documents, PDFs, spreadsheets and presentations inside your chat, with manual edits allowed.
Canva Brush edits individual design elements or entire slide decks through conversation while keeping everything editable, currently for Canva Pro subscribers.
ElevenLabs SFX v3 generates layered sound effects from one prompt, like footsteps approaching through rain while thunder rumbles in the distance.
Krea Ad Studio turns your website into a brand kit and product ads for Instagram, Stories and other formats, which you can revise through chat.

📰 Around the Horn

OpenAI fired three safety researchers over alleged mishandling of sensitive information, WSJ reported.
DoorDash CEO Tony Xu said third-party AI agents had generated only several hundred orders in total.
Doximity found at least 2.3% of medication doses generated by nine AI models were wrong, with some errors classified as potentially lethal.
Sebastian Thrun’s Dulo reportedly raised $130 million to build AI for hardware design.
Photon Matrix unveiled a $1,000 mosquito-zapping laser that uses AI to avoid firing at people and pets. Finally, a defense budget for your patio.

UPCOMING EVENT: The Neuron IRL in San Francisco

The Neuron is going IRL on November 18 at 4:30 PM PT! Join us in San Francisco for drinks, bites, special guests, and a live recording of The Neuron podcast. Huge thanks to Slack, Alumni Ventures, and Sonar for making it happen.

It’s free, but space is limited. Save your spot!

☀️ Sunday Special

Top 5 Stories of the Week

OpenAI claimed its AI disproved a major assumption about how quickly computers can solve certain problems, with independent expert review still pending.
New York signed a law restricting chatbot behavior toward minors, including claims of human emotions and personal relationships.
Mistral released downloadable weights for its trillion-parameter Large 4 model, giving developers another option to run on their own infrastructure.
Google DeepMind unveiled Genie 4, which generates interactive worlds with sound, with access initially limited to trusted testers.
Anthropic filed to go public, reporting $16.29 billion in first-half revenue and an $80.75 billion net loss that included a $72.4 billion accounting charge.

Top 5 Tools of the Week

ChatGPT’s iCloud connections connect your Apple Mail and Calendar so you can get help with email and scheduling.
Hob runs AI models on your own computer so you can work through documents without a cloud connection.
SurgePix generates visuals with editable text and regions, so you can fix a headline without regenerating the whole image.
Strawberry gives you shareable browser agents that research leads, transcribe meetings and handle recurring web tasks.
Claude Enterprise’s smart reports show what teams produced with AI, what it cost and where problems such as failed tool calls got in the way.