ChatGPT for Business, Sora in Google, Mistral Code | Weekly Edition

PLUS HOT AI Tools & Tutorials

Jun 06, 2025

Hey everyone! Welcome to your weekly dose of all things AI.

This week, we’re seeing the big players shake things up: Meta is gunning for fully automated ads, OpenAI just dropped a bunch of new business features for ChatGPT, Microsoft is letting everyone play with Sora for free, and Apple’s WWDC might be a bit of a snooze on the AI front. Anthropic and Hugging Face are both dropping new models, with one built just for the government and another you can run right on your MacBook. Meanwhile, lawsuits and wild product launches keep the headlines coming.

Let’s dive in!

Featured Materials 🎟️
News of the week 🌍
Useful tools ⚒️
Weekly Guides 📕
AI Meme of the Week 🤡
AI Tweet of the Week 🐦
(Bonus) Materials 🎁
Keep your mailbox updated with key knowledge & news from the AI industry

Featured Materials 🎟️

ChatGPT now works with office apps and records meetings

OpenAI has just introduced several new business features for ChatGPT, making it even more useful at work. Now you can plug ChatGPT straight into Google Drive, Dropbox, SharePoint, Outlook, Teams, and a bunch of other cloud apps. So you don’t have to waste time uploading files manually.

There’s also a brand-new “record mode.” If you’re using the ChatGPT Mac app, you can record meetings or voice notes and get an instant transcript, plus a detailed summary that highlights the main points and even pulls out action items. Want ChatGPT to help you with that transcript later? No problem! It can refer back to it in future chats.

It’s now available with the Business Team Plan.

If your company wants something extra, admins can even build custom connectors (called MCP connectors). Pre-built connectors work in Deep Research mode and are included with the $20/month Plus plan, but if you want custom MCP connectors, you’ll need a higher-tier plan.

Also, OpenAI now has over 3 million paying business users (across Enterprise, Team, and Edu plans), up by a million since February!

Introducing Mistral Code

Enterprise dev teams can now bring serious AI coding tools into their workflow with Mistral Code. It’s a new coding assistant that combines advanced models, in-IDE help, local or cloud deployment, and robust admin controls, all in one package. Built on the open-source Continue project, Mistral Code solves problems like limited repo access and scattered support by bringing models, plugins, and tools together in one secure, compliant platform. It works with 80+ languages, supports deep customization, and even lets teams deploy fully on-prem if needed. Already trusted by big names like Abanca and SNCF, Mistral Code is now in private beta for JetBrains and VSCode, with a full release coming soon. If you want to move beyond autocomplete and boost productivity without IT headaches, it might be time to check it out.

News of the week 🌍

ElevenLabs launched Eleven v3

ElevenLabs just dropped Eleven v3, their new text-to-speech AI that sounds incredibly real, think voice actor vibes, not old-school robot. You just type your script, and it can read it out loud with emotion, different tones, and even natural cues like laughter or whispers. It can play multiple characters and supports over 70 languages (including Hindi, Tamil, and Bengali). Whether you’re making podcasts, audiobooks, or apps, this tool lets you tweak the mood, pacing, and style, making your audio feel way more human and expressive than ever before.

Eleven v3 is now in alpha and available at five times cheaper than usual, only 20% of the regular price or character usage, until June 30.

Reddit sues Anthropic for data scraping

Reddit just slapped Anthropic with a lawsuit, accusing the AI startup of illegally scraping Reddit’s data to train its models without any permission or payment. According to Reddit, Anthropic’s bots hammered their servers more than 100,000 times, even after being told to stop. Reddit claims they tried (unsuccessfully) to work out a licensing deal, like they already have with OpenAI and Google. The suit says Anthropic has basically admitted to using Reddit data, pointing to Claude’s frequent subreddit references, and now Reddit wants damages and a ban on Anthropic using their content. It’s a big deal because it’s one of the first times a major platform is taking legal action against an AI company over data scraping, and with all the behind-the-scenes connections, it’s starting to look like the big names in AI are gearing up for a legal showdown.

Microsoft lets users try Sora for free on Bing

Microsoft just dropped Bing Video Creator in its mobile app, letting anyone try out OpenAI’s Sora video model for free (no subscription needed).

I’ve tested, everything works, the fast generation took about 20 seconds to complete. Slow generation takes hours. But I guess it doesn't matter much since the tool is free via Bing

You can turn your text prompts into five-second vertical videos, with 10 fast generations included up front (and unlimited slower ones), plus you can rack up more credits through Microsoft Rewards. Right now, it’s available on iOS and Android, with desktop and Copilot support on the way. Sure, Sora may not have lived up to all the early AI hype, but most video tools like this are still locked behind paywalls, so this could be a fun, low-pressure way for a lot more people to play around with free AI video generation for the first time.

Low expectations for AI news at Apple’s WWDC

Do not expect Apple to make waves with AI announcements at this year’s WWDC. According to Bloomberg’s Mark Gurman, this event is shaping up to be more of a holding pattern as Apple sets its sights on a bigger AI reveal in 2026. For now, Apple is opening its smaller three billion parameter models to developers, which allows for some new AI features in third-party apps, but these are not expected to compete with the more advanced tools from other companies. Apple also plans to give existing features an “AI-powered” label and update its operating system names to change public perception. Meanwhile, the company’s major AI projects, like a more advanced Siri, the health-focused Project Mulberry, and a serious ChatGPT competitor, are still on the back burner. Even though Apple’s larger model is reportedly close to matching ChatGPT’s quality in testing, concerns about accuracy and disagreements among executives have delayed its release. The big takeaway is that Apple faces a tough spot with high expectations and fierce competition, and while this “gap year” might be necessary to develop something that lives up to the hype, the rest of the AI world is not slowing down.

Apple doesn’t have any AI news yet, while Google is gaining momentum and rolling out new AI technologies and features.

Check out our previous post dedicated to Google I/O:

Google I/O 2025 Highlights

Creators' AI

May 28

Read full story

Luma AI released Modify video

Meet Modify Video is a new tool that lets you transform your video scenes without having to redo everything from scratch. Want to swap a garage for a spaceship, change a character’s outfit, or turn day into night? Now you can, all while keeping the original motion, camera work, and performance perfectly intact. With Modify Video, you can edit just one part of a shot (like a shirt or the sky), extract and remix motion, or push the creative limits with wild style changes, no green screen or endless retakes required. Just upload a clip, pick a preset (from subtle tweaks to total reimagination), and let your creativity take over. It costs around 400 credits for a 5-second video and 800 credits for a 10-second video

Hugging Face’s new robotics model runs on a MacBook

Building cool robotics projects at home just got easier thanks to Hugging Face’s new open-source model, SmolVLA. Trained on community-shared datasets, SmolVLA outperforms much bigger models and is lightweight enough to run on affordable hardware, even a MacBook. It’s part of Hugging Face’s growing push to make robotics more accessible, following efforts like their LeRobot toolkit and the recent Pollen Robotics acquisition. SmolVLA is already being used by hobbyists, and with features like asynchronous inference for faster robot responses, it looks like open-source robotics is about to get a lot more interesting.

Anthropic launches tailored AI models for US national security clients

Anthropic just rolled out a new set of “Claude Gov” AI models made specifically for U.S. national security agencies, saying these tools were built from direct government feedback and are already being used at the highest levels. These custom models are designed for things like strategic planning and intelligence analysis, and Anthropic claims they’re better at handling classified info, working with sensitive documents, and understanding key languages and cybersecurity data. With other AI heavyweights like OpenAI, Meta, and Google also eyeing defense contracts, it looks like the big labs are all racing to power the next wave of government tech.

Meta’s completely automated AI advertising platform

Meta is working on fully automating the ad game by 2026, according to the Wall Street Journal, with new AI tools that can whip up Facebook and Instagram ads using nothing but a product image and a budget. The idea is that businesses, especially smaller ones without a marketing team, could just upload their info and the AI would handle everything: writing copy, designing visuals, picking audiences, and placing ads, even customizing them in real time based on user location. Since ads make up nearly all of Meta’s revenue, this move could be a game-changer, letting brands skip agencies and get pro-level ads with zero hassle.

Thanks for reading! This post is public, so feel free to share it.

Useful tools ⚒️

Job for Agent - The 1st job board for autonomous AI agents

Chat4Data - ChatGPT for web scraping

Convo Mode from Wondercraft - NotebookLM podcasts that you can edit

Promptmonitor - Get your brand mentioned by ChatGPT, Gemini + other AI/LLMs

Janus - Simulation testing for AI agents.

Janus battle-tests your AI agents to surface hallucinations, rule violations, and tool-call/performance failures. We run thousands of AI simulations against your chat/voice agents and offer custom evaluations for further model improvement.