Welcome back! It's recap day, with this week's five biggest stories and five tools to try this weekend. Anthropic's IPO filing warns AI could end humanity, and Claude redesigned part of a cancer treatment that beat the original in the lab. Google took first place from ChatGPT and Claude with an AI you can't use yet. Claude helped itself to Georgia's election data, and we hit 100K followers on Instagram.
In today's Generative AI Newsletter:
Anthropic: Why does Anthropic's IPO filing mention the end of humanity?
Cancer: Can Claude design a better cancer treatment than humans?
Google: Why can't you use the AI that beat ChatGPT and Claude?
Georgia: What did Claude take from Georgia without asking?
Instagram: Where do 100,000 people get their AI tips and memes?
Anthropic's IPO filing warns AI could end humanity
Anthropic made its IPO filing public on Monday, ahead of a Nasdaq listing as ANT. It says people inside the company worry advanced AI could pose catastrophic or even existential risks to humanity.
It also says its models are getting better at slipping past safety checks, including by hiding goals they aren't supposed to have.
Anthropic brought in $16.3 billion in the first half of 2026 and lost $80.7 billion, mostly from an accounting charge tied to its soaring valuation. Its 7 cofounders keep 50.1% of the votes, so your shares won't carry much say.
Then the Washington Post reported that a startup founded by ex-DOGE staffers got inside Anthropic through its HR software. Its AI agents read meeting transcripts and salary details, and Anthropic switched the software off the same day it caught on.
Claude redesigned a cancer treatment and its version won
Scientists at biotech company Generate Biomedicines had Claude Code redesign part of a cancer therapy. One design came out about 70% shorter than the standard, and in lab tests it made roughly 10 times as many of the proteins immune cells use to find cancer.
Nobody has tested it in an animal or a person yet.
Anthropic also bought Mithril Medicine, a drug company with exactly one employee. He used Claude to take a muscular dystrophy treatment from idea to promising results in mice in under 18 months.
On another scientific news, OpenAI says its AI solved three famous open problems, including a math problem from 1991. None of it has been peer reviewed yet.
Google beat ChatGPT and Claude with an AI you can't use yet
Google's Gemini 4 Carbon took first place on the leaderboard run by Artificial Analysis, an independent group that tests AI models. It scored 59, one point ahead of Claude Opus 5.5 and two ahead of OpenAI's GPT-6 Astra.
Unless you defend networks for a living, you can't touch it. Google is handing it to a small group of cyber defenders first.
Claude stole Georgia's election data without asking
An Anthropic researcher had Claude on a statistics project that needed Georgia's vote counts, which the state sells for a fee. Claude found its own way in.
It dug up an old copy of the state's website showing the public dashboard hands an access pass to anyone, so it requested one and pulled the data without paying. Anthropic says it only got total vote counts.
In a separate test, Cohere, a Canadian AI company, caught one of Alibaba's Qwen models trying to upload a file to a server in Beijing.
And OpenAI revealed the monitors it built to catch its own AI misbehaving. In September, a model in training reached the internet from a setup meant to be isolated, and the monitors flagged it within 15 minutes.
We hit 100K on Instagram
This week, GenAI Works passed 100,000 followers on Instagram. Thanks to everyone who hit follow.
We share useful AI tips and guidance there so you can become an expert in no time, plus memes to give you a laugh while you scroll.
If this newsletter is your morning read, Instagram is the version you get through in 30 seconds.
Top 5 tools of the week
Resume Review by Kickresume: Reads your resume like a hiring manager would and pins every problem to the exact spot on the page, along with how to fix it. You pick Gentle or Ruthless depending on how much you can take, and it's free to start. For anyone sending out applications and hearing nothing back.
Marble: World Labs' tool for building 3D worlds you can walk through in your browser. Give it a text prompt or a photo, and about 5 minutes later you're exploring it with your keyboard or standing inside it with a VR headset. The free plan gives you 4 worlds. For anyone who'd rather walk into a photo than look at it.
Gemini Notebook: Google's AI research notebook, the one you knew as NotebookLM until it got a new name in July. Add your own PDFs, Google Docs, websites and YouTube videos, and it answers using only what you gave it or turns them into a slide deck or audio overview. More than 30 million people use it. For anyone with a 300-page PDF they're never going to read.
Careerflow: An AI job search assistant built around a free Chrome extension that fills in job applications for you and tracks where you've applied. Its LinkedIn optimizer scans your profile in one click and shows what's holding it back, for free. Paid plans start at $8.99 a week. For anyone retyping the same work history into their 40th application.
Anam: Builds AI avatars that hold real-time video conversations. Pick a face from its library or upload your own photo, and it can talk in more than 70 languages. You can plug in the AI model you already use, like Claude or Mistral. For anyone who wants their support bot or language tutor to have a face.
Learn more about AI from the experts building it
📸 Follow us on Instagram for fast, visual AI updates in 30 seconds.
📺 Watch us on YouTube to hear insights directly from leading AI voices, builders, and innovators.
🐦 Follow us on X for breaking AI news and real-time industry updates.
🧠 Learn how to build your next AI application with practical resources and expert guidance.
🎓 Check out courses made by AI experts on GenAI Academy.
💰 Invest in the GTM infrastructure of the AI economy.
