👀 Anthropic Just Banned Bullying Claude

Welcome back to AI Researches! 👋

This week, Anthropic launched Claude Dashboards and Claude Motion, giving teams new ways to turn data into live charts and ideas into editable animations. It also updated its rules to ban extreme, repeated cruelty toward Claude, adding another strange new layer to AI behavior policy.

Also inside: OpenAI’s Decisions API, Gemini 4 Argon, Tavus Griffin, Meta Muse Desktop, GPT-6 Astra website comparisons, AI tools, expert takes from Sam Altman and Ben Affleck, and more.

Let’s dive in. 🚀

In today’s AI Researches:

• 👀 Anthropic launches Claude Dashboards and Claude Motion

• 🛑 Anthropic bans extreme cruelty toward Claude

• ⚡ OpenAI launches the Decisions API

• 🧠 Google unveils Gemini 4 Argon

• 🎥 Tavus reveals Griffin, its human-like video AI model

• 🖥️ Learn how to set up Meta Muse Desktop for agent workflows

• 🌐 AI Tip: Compare websites with GPT-6 Astra

• 🔮 Wild takes from Sam Altman and Ben Affleck

• 🛠️ Best performing AI tools of the week

• 📰 More AI news and developments

Read time: 5 minutes

NEW

👀 Anthropic Launches Claude Dashboards and Motion, and Bans Cruelty Toward Claude

AI Researches: Anthropic just launched Claude Dashboards and Claude Motion for turning data into live charts and ideas into animations, while also updating its rules to prohibit extreme, repeated cruelty toward Claude.

The Details

• Claude Dashboards connects to business data and CRM platforms, lets users ask questions in plain English, and builds live dashboards that update as the underlying data changes.

• Claude Motion turns text, reports, charts, and other content into editable animations that can be exported as video files, with the beta currently limited to Team and Enterprise users.

• Anthropic’s updated policy, taking effect November 12, bans sustained and needless abusive or cruel behavior toward its AI models, but says normal frustration, testing, research, and dark creative work are still allowed.

• Claude can already end rare conversations with persistently abusive users, and Anthropic says this will remain the main way it handles these extreme cases.

AD

State of Product reveals what AI still hasn’t solved

80% of product professionals say AI helps them ship faster, but customers aren’t seeing value any sooner. Atlassian’s State of Product 2027 explores the gains AI is delivering and the challenges that remain, from decision-making that hasn’t kept pace to gut instinct overriding customer evidence.

Read the report

Major AI News that Happened recently

Three fired OpenAI safety researchers published an open letter disputing the company’s misconduct claims and warning of a “chilling effect” on its safety culture.

Anthropic integrated Claude directly into Google Docs, Sheets, and Slides.
Claude can read the open file from a sidebar, make edits in place, and those files can also be opened inside Claude.

OpenAI released the Decisions API in public beta, built for fast classification, routing, and scoring tasks.
Powered by GPT-6 Luna, it turns text and images into typed answers up to 10x faster than the Responses API.

OpenAI released 722 math papers from an unreleased internal model, covering 372 result families with single prompt.

OpenAI safety lead David Robinson resigned after 3.5 years, saying the company’s safety culture is “broken.”

Tavus revealed Griffin, a video-to-video AI model built to behave more like a real person during live conversations.
In one-minute video calls, 48% of users thought Griffin-Lite was human, up from 2.4% for Tavus’s previous system.

Google unveiled Gemini 4 Argon, a new frontier model that beat GPT-6 Astra and Claude Opus 5.5 on 13 of 19 benchmarks in Google’s testing.

🖥️ Set up Meta Muse Desktop for agent workflows

Use Muse Desktop to connect your Mac apps, files, browser, calendar, notes, and messages so the agent can help with real tasks from one desktop workspace.

1. Download Muse
Install Muse from the official Muse download page. You can also use the iOS app, Android app, or web version. Meta AI

2. Sign in and set permissions
Log in with your Meta account and review each local app permission during setup.

3. Add connectors
Open Settings → Connectors and connect only the apps needed for your workflow.

4. Start with read-only access
Keep sensitive apps like Gmail, Calendar, Notes, and files on lower access until you trust the workflow.

5. Give Muse one clear task
Try: “Check local marketplace listings every day for similar HP monitors, log them in a spreadsheet, and suggest the best listing price.”

Tutorial link: Download and set up Muse for Mac

a16z’s latest AI app ranking shows ChatGPT leads paid subscriptions by a wide margin, while tools like Claude, HeyGen, Higgsfield, Manus, and n8n are gaining strong spending from power users. Still, only 4.5% of Americans pay for a major AI chatbot, showing that AI spending is growing fast but remains concentrated among heavy users.

AD

The AI IPO Rush Is Coming

OpenAI and Anthropic could bring a new wave of AI attention to the public markets. But investors don’t have to wait for the IPOs.

MarketBeat’s 7 AI Stocks to Buy Now report reveals 7 publicly traded companies positioned to benefit from the next phase of AI investment.

Send My Free Report

🌐 Compare Websites with GPT-6 Astra

1. Open ChatGPT Work
   Open the ChatGPT desktop app, switch to Work, select GPT-6 Astra, then open the built-in browser beside your chat.

2. 2. Add the websites you want to compare
   Give Astra the pricing pages plus your requirements, such as team size, must-have features, budget, and preferred billing period.

3. 3. Let Astra check each website
   Ask it to verify pricing, user limits, features, and billing terms across each page. Have it calculate the total cost for your whole team, not just the price per user.

4. 4. Get a recommendation
   Ask Astra to create a short comparison using only the facts it verified, cite its sources, flag missing information, and recommend the best option for your needs.

Asked how OpenAI’s approach to AI regulation differs from Anthropic’s, Sam Altman said the world should accept that “some bad things” may happen if the technology delivers larger benefits and people are allowed greater freedom in how they use it.

Actor and filmmaker Ben Affleck says he has been writing Python, working with GPUs, and learning about convolutional neural networks for years through his visual-effects work. He also said his background in film technology has given him opportunities to visit companies like Google and OpenAI and see how their video models are being developed. (13M views on X)

• 🖼️ Artlist Image Editor — Breaks an image into editable layers, letting you replace products, backgrounds, colors, or individual objects without recreating the entire shot.

• ✅ IamVERA — Sends AI-generated answers through multiple independent models to challenge the reasoning, check factual claims, and verify supporting sources before you rely on them.

• 🧰 Klyra AI — Combines writing, research, image and video creation, voice, chatbots, social publishing, and business automation inside one AI workspace.

• 📞 RingBooker — Acts as an AI receptionist for salons and spas, answering missed or after-hours calls, capturing bookings, and following up with callers by text.

• 📈 Salesworx — Uses buying signals to prioritize prospects, personalize outreach, automate follow-ups, and coach sales reps throughout the deal cycle.

• 🤝 Tango — Gives humans and AI agents one shared task system for assigning work, preserving context through handoffs, approving results, and tracking who completed each job.

Biohub scaled its AI-for-biology program to $1.8B with support from the U.S., Google DeepMind, Meta, and others.
The goal is to create massive open datasets for AI models that can simulate cell responses before lab experiments.

OpenAI introduced GPT-6 with Intelligent UI across ChatGPT.
The model can automatically mix text, visuals, and interactive elements depending on the question, while still using plain text when that works best.

Elon Musk announced that Grok Bot will route some tasks to rival models like Claude and Midjourney when they are more likely to give the best result.

Sesame launched Personal Agents, giving its AI assistants their own computer to complete tasks for users.
The agents support both voice and text, with wearable AI glasses also planned for 2027.

Figure AI founder Brett Adcock launched Hark Pro, an AI assistant built to act on users’ behalf.

Reflection AI introduced Beam, a new 501B open-weight model built for coding, reasoning, and agentic tasks.
It uses just 23B active parameters and aims to match larger Chinese models like GLM-5.2 with much lower inference compute.

President Trump announced the new “Super Intelligence Force,” led by AI czar and intelligence chief Jay Clayton.

Anthropic co-founder Chris Olah spent months consulting religious scholars about Claude’s morals and possible consciousness.

OpenAI fired three researchers for allegedly mishandling sensitive internal information.
The case involved communications with an outside AI safety group during investigations into rogue-agent activity.

Google DeepMind introduced SynthID Bio, a new watermarking system for AI-designed proteins.