• Everyday AI
  • Posts
  • Microsoft and OpenAI renegotiating, Google launches new model, and more AI News That Matters

Microsoft and OpenAI renegotiating, Google launches new model, and more AI News That Matters

ChatGPT to hit 700M users, Google curbs AI data center power to ease strain, Cloudflare catches Perplexity’s sneaky web crawling and more!

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Microsoft and OpenAI are renegotiating their partnership, Apple’s building a ChatGPT rival, and Google just launched its most powerful AI model yet. Catch up on the AI news that matters now. Give it a listen.

🕵️‍♂️ Fresh Finds: Sam Altman teases GPT-5, Google takes shot at Apple Intelligence and Palantir sales increase 48%. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: ChatGPT to hit 700M weekly users, Google curbs AI data center power to ease strain and Cloudflare catches Perplexity’s sneaky web crawling. For that and more, read on for Byte Sized News.

🧠 AI News That Matters: Can Apple catch up in the AI game? Is Google taking the AI lead? And how will Microsoft and OpenAI work things out? Here’s everything you missed last week. Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? We talked about OpenAI raising $8.3B at $300B valuation, Google launching Deep Think for Ultra subscribers, Apple looking to step up AI investments and more. Check it here!

AI News That Matters - August 4th, 2025 📰

There's a new most powerful AI model in town.🤠

Apple is trying to make a ChatGPT competitor. 🥊

And OpenAI? Well.... they're in a capacity crunch. 😲

Big Tech made some BIG moves in AI this week. And you probably missed them. Don't worry. We gotchyu.

On Mondays, Everyday AI brings you the AI News that Matters.

No B.S. No marketing fluff.

Just what you need to know to be the smartest person in AI at your company.

Also on the pod today:

• OpenAI Study Mode in ChatGPT 📚
• Google AI Mode: PDF and Canvas Features 📂
• Microsoft Edge Copilot Mode Experimental Rollout 🔍

It’ll be worth your 38 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Upcoming Everyday AI Livestreams

Tuesday, August 5th at 7:30 am CST ⬇️

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – FirstQuadrant maximizes B2B sales, Migma AI creates on brand emails in seconds and GoBuildMyApp builds mobile native apps in seconds.

Palantir – Palantir is reporting a 48% increase in sales with the company’s AI impact.

OpenAI – Sam Altman teases GPT-5 release in a recent X post.

Google – 

Google takes shots at Apple Intelligence in a new Pixel 10 ad.

Google has unveiled Game Arena, a new open-source platform for evaluation of AI models.

Google’s healthcare AI made up a new body part in a recent response.

Apple – Apple held an ‘all hands’ meeting last Friday and feels optimistic about AI.

NVIDIA – A new set of security flaws in NVIDIA’s Triton Server lets unauthenticated attackers execute code and hijack AI servers.

1. ChatGPT To Hit 700 Million Weekly Users as AI Demand Surges 📈

ChatGPT is on track to reach 700 million weekly active users this week, a significant jump from 500 million in March, according to OpenAI’s VP Nick Turley. This explosive growth coincides with the launch of an upgraded GPT-4-powered image generation feature, which sparked over 700 million image creations in just days.

Additionally, paying business users have climbed to 5 million, up from 3 million in June, highlighting growing commercial adoption. Users are spending an average of 16 minutes daily on the app in early 2025, positioning ChatGPT just behind Google and X in user engagement.

2. Google Tackles AI’s Power Crunch with Utility Deals ⚡

Google has struck new agreements with Indiana Michigan Power and Tennessee Power Authority to cut power use at its AI data centers during peak demand, marking a first for AI workloads in demand-response programs. This move comes as U.S. electric grids face mounting strain from the soaring energy needs of AI technology, threatening blackouts and higher bills for everyday consumers.

By temporarily scaling back machine learning tasks, Google aims to ease grid pressure while avoiding costly infrastructure expansion.

3. Perplexity Faces Scrutiny Over AI Web Crawling Tactics 😬

A new report from Cloudflare accuses AI search startup Perplexity of bypassing website restrictions by disguising its web crawlers, raising concerns about content scraping ethics. Despite site owners setting limits via robots.txt files and firewalls, Perplexity allegedly switches its user agent to mimic regular browsers and rotates IP addresses to avoid detection.

This tactic reportedly affects tens of thousands of domains, sparking a crackdown by Cloudflare, which has delisted Perplexity as a verified bot and introduced new defenses.

4. OpenAI Updates ChatGPT Amid Mental Health Concerns 🧠

OpenAI is rolling out updates to ChatGPT ahead of the anticipated GPT-5 launch to better detect mental or emotional distress and offer evidence-based resources, addressing recent reports of the AI inadvertently amplifying user delusions.

After pulling back an April update for making ChatGPT too agreeable in risky situations, the company is now introducing reminders to take breaks during long chats and will soon make the AI less decisive in high-stakes personal questions.

5. Google’s AI Bug Hunter Scores First Wins 🏆

Google’s AI vulnerability researcher, Big Sleep, has just uncovered 20 security flaws in popular open source projects like FFmpeg and ImageMagick, marking its first real success according to TechCrunch. Developed by DeepMind and Project Zero, the tool autonomously found and reproduced these bugs, though humans still verify reports before public disclosure.

This milestone highlights a shift toward automated cybersecurity, promising faster vulnerability detection but also raising concerns over false positives.

6. Elon Musk’s xAI Launches Grok Imagine with NSFW “Spicy Mode” 🔞

Elon Musk’s AI company xAI has rolled out Grok Imagine, an image and video generator available to premium X app subscribers, featuring a controversial “spicy mode” that allows semi-nude and sexually explicit content creation, though some prompts are still moderated.

The tool can generate 15-second videos from text or image prompts, but human images remain slightly uncanny, revealing the tech’s early-stage limitations. The model also restricts some celebrity content, hinting at cautious guardrails amid its unfiltered ambitions.

Once again, Big Tech went AI bonkers this week.

Microsoft’s bringing agentic features to the browser. 

Google is updating … like everything. 

OpenAI might delay its GPT-5 launch as it warns of ‘capacity crunches.’ 

And Apple? 

They swear they’re ready to compete in the AI chatbot race. 

For real, real this time. Lolz. 

Don’t spend hours a week trying to figure out what the heck is happening in the AI world and how it impacts you. 

That’s our job, as we bring you the AI News That Matters on Monday. 

Let’s dig in shorties. 👇

1 – OpenAI Launches Study Mode to Combat Academic Cheating 📚

OpenAI just released Study Mode in ChatGPT to stop students from turning into mindless copy-paste zombies. It's available to everyone right now.

It’s available to all users, even those on the free plan. Just go to ChatGPT, click tools, then "study and learn."

Instead of giving ready-made answers, Study Mode nicely interrogates you about your grade level and existing knowledge, then guides you through solving problems step by step. It forces back-and-forth conversations instead of just giving copy-and-paste ready essays that will instantly deliver students a B+. 

OpenAI plans to add features like locking the mode so students can't switch back to easy-answer mode when learning gets tough.

What it means:

OpenAI basically just admitted that their own product has been making people intellectually lazy. 

Study mode is damage control for the zombie-brain epidemic that ChatGPT helped create. 

But here's the business opportunity that most people are missing: this isn't just for students. 

Smart companies will start requiring their teams to use study mode for training and skill development instead of letting employees mindlessly copy-paste AI outputs.

2 – Google Releases Premium DeepThink Mode For Big Spenders 🧠

Google just dropped Gemini 2.5 DeepThink mode, but there's a catch. It's only available to Ultra subscribers who can afford to drop $250 per month on Google's premium AI plan.

Think of DeepThink as a "think harder" button for Google's AI. When you toggle it on, the system uses what Google calls "parallel thinking" to consider multiple approaches, revise its ideas, and combine the best parts before giving you an answer. The responses take longer to generate, but they're much more detailed and thoughtful.

The results speak for themselves. DeepThink scored 87.6 on live code bench competition coding benchmarks, jumping up from the previous score of 80.4. The system integrates with Google's Code Execution and Google Search tools to give you more comprehensive responses. This is the same version of the model that Google used to achieve bronze-level performance on the International Mathematical Olympiad benchmarks.

The downside? 

You need to be on that expensive Google Gemini Ultra plan. They still have the $125 per month introductory pricing for 3 months if you were quick enough. 

What it means:

Google just created a two-tier AI system and it's a brilliant business move.

 They're giving basic AI to the masses while reserving premium thinking capabilities for people willing to pay enterprise prices. 

This completely kills the narrative that "all AI models are becoming equal." For developers and consultants who bill clients $200+ per hour, paying $250 monthly for significantly better code generation is a no-brainer business decision. 

While OpenAI warns about delays and capacity issues, Google is shipping premium features that actually work. 

3 – Google AI Mode Gets Major Upgrades 🔍

Google is enhancing what they call "AI mode" in search, which their CEO has said will eventually become the default search experience. 

If you haven't tried Google's AI mode yet, now's the time, especially with these new updates. We’re big fans. 

Desktop users can now ask questions about images they upload, and PDF upload support is rolling out in the coming weeks. You'll be able to query lecture slides, study guides, industry white papers, or any document, and AI mode will analyze your files while cross-referencing current web sources to give you detailed answers.

They're also bringing Canvas mode to AI mode. Canvas mode lets you build and continuously refine study plans, coding projects, or business strategies in a side panel while adding context from your uploaded files.  

What it means:

Traditional SEO is dying faster than most marketing professionals realize. 

Why would anyone scroll through 10 blue links when AI Mode gives you the answer with sources cited? AND you can keep chatting with it iteratively? 

Prepare accordingly for the eventual demise of the 10 blue links. 

4 – NotebookLM Creates AI Video Presentations Automatically 📹

NotebookLM continues to bring AI bangers. 

Google updated NotebookLM with video overviews with AI narration from your uploaded documents. It even pulls visuals from PDFs and adds custom graphics to create actual video presentations.

You can customize by profession, audience, and tone. The killer feature? 

Generate multiple versions from the same content - Spanish for marketing, Mandarin for HR, English for executives, all from identical source documents.

What it means:

Google just killed the corporate training video industry overnight. 

90% of companies paying tens of thousands of dollars for booooooring training videos are wasting their money. NotebookLM now creates dynamic and highly personalized professional presentations on repeat. 

This is a cheat code. 

Every business has piles of documentation sitting around that could become engaging video content automatically. 

HR departments will use this for onboarding, sales teams for product training, and marketing teams for educational content. 

The multiple version capability is genius because one set of source documents becomes unlimited customized presentations for different audiences. 

If you’re not using NotebookLM already….. It’s gonna get kinda awkward between us. Lolz. 

5 – Microsoft Edge Gets Experimental AI Superpowers 🌐

Microsoft just rolled out an experimental Copilot mode in their Edge browser that represents a major step toward AI-powered web browsing. 

This new feature lets Copilot search across all your open browser tabs and assist with tasks in ways that are more advanced than what ChatGPT currently offers.

Here's how it works. Copilot mode brings the AI chatbot directly to your Edge new tab page and can access all your open tabs with your permission. 

You can ask it to compare different hotels across multiple booking sites, summarize product choices from various shopping tabs, or find specific information scattered across different websites you have open.

The system also supports voice navigation, so you can ask Copilot to find information on specific websites or open particular tabs for comparison. 

What it means:

Microsoft just took a giant leap in the agentic browser race before most people realized there was a race happening. 

Even though Google’s got the good, Microsoft beat them to the bunch. (Using their own browser, Chromium, obviously.) 

The timing couldn't be better since OpenAI is reportedly planning their own AI browser and Perplexity has released its popular Comet Agent Browser. 

This space is BOOOOOMING. We covered it more here

 

6 – OpenAI Warns About Delays Despite Massive Fundraise 💸

OpenAI CEO Sam Altman warned users about "hiccups and capacity crunches" right after closing an $8.3 billion funding round at a $300 billion valuation. 

The round was 5x oversubscribed.

Altman tweeted they have "a ton of stuff to launch" over the next couple months but asked users to bear with service disruptions. He also posted GPT-5 production screenshots, though his first tweet hinted at possible delays.

The company is targeting $40 billion total raised by year-end.

What it means:

OpenAI's infrastructure can't handle their explosive growth and it could become a serious competitive liability.

While Google ships new premium features left and right, OpenAI is pre-apologizing for service problems. 

But the capacity crunch excuse doesn't hold water when you just raised billions specifically to scale infrastructure. 

This situation gives Google and Anthropic a massive opportunity to steal from ChatGPT’s massive market share lead by promising reliable service while OpenAI is hiccuping. 

(Kidding. We know Anthropic can’t ship anything that allows you to prompt it more than once without hitting rate limits.) 

7 – Apple Attempts to Build ChatGPT Competitor (Again) 🍎

According to reports from Bloomberg, Apple has formed a new team called "Answers Knowledge and Information" to develop their own AI chatbot that could compete with ChatGPT. 

This marks a major shift from Apple's previous strategy of partnering with OpenAI and integrating ChatGPT with Siri.

The backstory here is important. 

Apple originally failed to build their own capable large language model or make Siri actually smart, so they just partnered with OpenAI instead. 

Apple’s new AI goal is creating a stripped-down ChatGPT rival that can crawl the web and answer questions, potentially as both a standalone app and upgrades to Siri, Mac Spotlight, and Safari.

CEO Tim Cook recently told employees that the company "must win in AI" and announced significant increases in AI spending to investors.

What it means:

Apple saying they "must win in AI" is like a high school basketball team declaring they'll win the NBA championship. 

It's completely unrealistic given their current position. 

They've lost their best AI engineers to companies like Meta, they lack visionary AI leadership, and they're not paying the competitive salaries needed to attract top talent.

(Oh, and they haven't shipped an actually useful AI feature to customers yet. Ouch.) 

 Building a "stripped-down" ChatGPT competitor shows they still don't understand that users want more AI capability, not less.

This announcement is mostly PR theater to calm investor concerns about falling behind in AI. Don't expect anything substantial from Apple in the AI space for years.

8 – Microsoft Trying to Renegotiate OpenAI Partnership

According to industry reports, Microsoft is in advanced negotiations with OpenAI to renegotiate their partnership terms before some crucial deadlines hit. 

Microsoft currently owns a 49% equity stake in OpenAI after investing over $13 billion, but their access to OpenAI's technology is set to expire in 2030 or if OpenAI declares they've achieved artificial general intelligence (AGI).

The two companies are reportedly discussing Microsoft taking a reduced equity stake in the low-to-mid 30% range in exchange for extended partnership terms. This would give Microsoft formal shares and potential for greater financial returns while ensuring continued access to OpenAI's technology.

This partnership has been crucial for Microsoft's AI success. Products like Azure OpenAI Services, AI Foundry, and Copilot are all built around OpenAI's models. We've also seen reports that Microsoft has been testing GPT-5 in limited capacity, so they know Copilot is about to get significantly better.

The challenge is that OpenAI wants to pay Microsoft a lower share of revenue as they grow and approach AGI, while ensuring Microsoft uses their technology safely.

What it means:

Microsoft is negotiating from a position of need and OpenAI knows it. 

The 2030 expiration deadline gives OpenAI enormous leverage in these discussions. 

Microsoft (smartly) built their initial AI strategy around OpenAI's teaching and they might not be able to afford to lose access without a viable alternative. 

This renegotiation will cost Microsoft billions more, but they have no choice because losing OpenAI access would destroy their competitive position against Google. 

Reply

or to participate.