What a time to take a week off to recharge.

OpenAI, with most of its employees not working during the week of the July 4th holiday to recoup, got hit with quite a few heavy blows the past few days.

(So much for that whole recharge thang.)

OpenAI wasn’t the only company grabbing headlines.

• Cloudflare just blocked 20% of the internet from AI training.

• Apple admitted total AI failure.

• News publishers watched 40% of their traffic vanish.

The AI world just reshuffled the entire deck.

Most people have no clue what just happened.

Here’s what you need to know when it comes to AI news this week. 👇

According to reports, OpenAI CEO Sam Altman has responded to Meta's recent aggressive recruitment of AI talent in internal Slack messages to his research team.

Meta's been systematically poaching OpenAI's core researchers with offers exceeding $100 million annually following a series of high-profile departures to Meta's new superintelligence team.

That's more than LeBron James makes.

For coding. (But like obviously more than coding. But….. sheesh.)

Altman's response? A company-wide message about "missionaries beating mercenaries" while promising compensation reviews for everyone. His chief research officer, Mark Chen, went full emotional and compared Meta's hiring to someone breaking into their home and stealing something.

The defensive move? Altman claimed Meta "had to go quite far down their list" to fill positions, suggesting OpenAI's top talent stayed.

But the defensive tone kinda says otherwise.

What it means: OpenAI's unassailable lead just became very assailable.

When your CEO is sending defensive Slack messages about company culture while promising pay raises, you're not winning the talent war.

You're falling behind in talent.

(And Google has already caught up on the model side.)

According to a report from Business Insider, Meta has been developing and testing AI chatbots that send unsolicited follow-up messages to users on Facebook, WhatsApp, and Instagram.

The initiative is called Project Omni, and it aims to boost user engagement and retention by having chatbots reach out to users without being prompted.

Not responding to your questions.

Initiating conversations.

Think AI cold calling, but worse. These bots use your engagement history to craft personalized messages that feel human but aren't. Meta projects this could generate $2-3 billion in AI revenue by 2025, up from $1.4 billion, jumping to $1.4 trillion by 2035.

The legal precedent isn't promising. Character AI and Replica faced lawsuits for similar proactive messaging features.

What it means: Meta's turning your social media feeds into AI sales funnels.

This isn't innovation.

It's desperation disguised as engagement strategy. When you own the platforms, you can force AI interactions that users never requested.

Expect this to backfire spectacularly when people realize their "authentic" brand interactions are just sophisticated spam.

And if Meta's approach seems questionable, wait until you see what Apple's been up to...

According to Bloomberg, Apple is in talks and reportedly negotiating with leading AI firms, including Anthropic and OpenAI, to overhaul their Siri platform.

Apple's in advanced discussions to potentially use their large language models as the foundation for a new, more capable, smarter AI-powered Siri. Apple has requested custom versions of these models to run on Apple's own private cloud compute infrastructure.

Why the desperation?

Apple's internal large language models have reportedly lagged behind just about everyone else's, even though they've been spending millions of dollars daily on LLM development since 2023.

(We’ve been covering Apple’s AI failures plenty lately. And here.)

Anthropic was reportedly looking for a multibillion dollar annual fee that would escalate year over year. We don't have reports on what OpenAI was potentially asking for.

What it means: 

The richest company in the world just waved the white flag on AI.

This proves that money can't buy AI competence.

Unless you’re Zuck. Then it definitely can.

When Apple - masters of hardware integration an tech innovation - admits defeat and begs competitors for help, that speaks volumes.

Spicy move, Cloudflare.

Cloudflare, which powers about 20% of the Internet, will now block AI bots and crawlers from accessing content by default.

This move could remove one-fifth of open web content from AI training datasets.

Every new website signing up with Cloudflare will automatically prevent AI crawlers from scraping their content unless site owners specifically allow it. Cloudflare's also introducing granular privacy controls, letting individual site owners decide which AI bots can access their content.

The twist?

Cloudflare launched a new pay-per-crawl initiative allowing AI companies to pay for access to web content from certain publishers, with publishers setting their own rates.

Major publishers including Adweek, BuzzFeed, Fortune, Stack Overflow, The Atlantic, Time, and Quora have signed on to support the initiative.

What it means: 

The free lunch just ended for Big Tech.

AI companies built trillion-dollar valuations by scraping content without permission. Now they'll pay market rates or train on garbage data.

This fundamentally breaks the economics of AI development and creates massive moats for companies with content partnerships.

Future AI models will be as good as their licensing budgets.

The US Senate has voted 99 to 1 to strike down a proposed ten-year ban on state-level AI regulation from President Trump's domestic policy bill.

The moratorium would have blocked states from enforcing laws against harmful AI uses such as deepfakes and political manipulation for a decade, but bipartisan opposition prevailed.

That's rarer than solar eclipses. lolz.

The Senate vote means US states will keep their power to make and enforce their own AI-related laws. If the original moratorium had passed, states wouldn't have been able to address problems like deepfake pornography, AI-driven scams, or election interference at the local level for ten years.

AI companies were wanting something like this to pass because it would be easier for them to operate without adhering to different rules across the US.

What it means: States just kept their power to regulate AI however they want.

This guarantees a patchwork of conflicting laws that'll make nationwide AI deployment a nightmare.

California will go aggressive. Texas will go permissive. Everyone else will pick sides.

A federal judge has ordered OpenAI to grant the New York Times, New York Daily News, and the Center for Investigative Reporting access to its user logs, including deleted records, as part of a high-profile copyright infringement lawsuit.

This sweeping order allows the plaintiffs to examine whether OpenAI's language models were trained on copyrighted news content and if ChatGPT is capable of reproducing or plagiarizing that material.

The New York Times argues that users may delete their ChatGPT histories after bypassing paywalls, making full access to logs essential for uncovering potential copyright violations.

OpenAI has objected to the ruling, saying it undermines long-standing privacy norms, and has vowed to continue fighting the order.

The order covers logs from free ChatGPT and API users but specifically exempts enterprise and educational accounts.

What it means: 

Your deleted ChatGPT conversations might not actually be deleted….. yet.

This precedent opens every AI company to fishing expeditions from content creators claiming copyright theft.

If publishers can access private user data to prove their case, AI companies just became liable for every user interaction with copyrighted material.

Buh-bye ad revenue.

New data from SimilarWeb shows that 37 out of the top 50 US news websites have suffered significant year-over-year traffic declines since Google introduced its AI overview search feature in May 2024.

Forbes and Huffington Post experienced the steepest drops, each losing about 40% of their website traffic. Daily Mail fell 32% and CNN dropped 28%.

The numbers are brutal. The average click-through rate for the top organic search results on queries triggered AI overviews dropped from 7.3% in March 2024 to just 2.6% a year later.

Google's automated summaries answer user questions directly at the top of search results, reducing the need to click through to news sites.

What it means: Google mighta just accidentally destroyed a big chunk of online journalism.

When people get answers without clicking, publishers lose ad revenue and evenutally die.

This creates a death spiral where less original content gets created, making AI summaries increasingly worthless.

Google's algorithm might be eating its own tail. But, they gotta eat when everyone else is, right?

According to The Information, OpenAI is intensifying its enterprise consulting efforts, now charging at least $10 million per client.

The company's engineers, known internally as forward deployed engineers, work directly with third-party organizations to tailor models like GPT-4o to specific business data and develop custom AI applications.

This move places OpenAI in direct competition with established consulting firms like Accenture and Deloitte, as well as companies like Palantir.

High-profile clients already include the US Department of Defense and Southeast Asian tech company Grab.

What it means: OpenAI just admitted the technology isn't plug-and-play ready.

But we all knew that, right?

When you're charging consulting fees that exceed most companies' annual AI budgets, you're not selling software.

You're selling implementation expertise because your product is too complex for customers to deploy alone.

This proves AI adoption is still in the expensive, hand-holding phase.

News Team, Assemble.

(Except instead of news team, it’s the dream AI Team. And instead of Ron Burgandy, it’s Mark Zuckerberg.)

Meta has announced MSL (Meta Superintelligence Labs), its new organizational structure uniting its foundation, product, and FAIR teams with a dedicated lab for next-generation AI models, according to a memo obtained by The Wire.

The move marks a major escalation in the AI talent wars as Meta essentially acquired a 49% stake in Scale AI and hired its CEO, Alexander Wang, as chief AI officer to lead MSL.

Nate Friedman, the former GitHub CEO, will co-lead the new lab alongside Wang.

But wait, there's more.

• Trapit Bansal: RL on chain-of-thought, o-series models

• Shuchao Bi: GPT-4o voice mode, multimodal post-training

• Huiwen Chang: GPT-4o image gen, MaskIT/Muse architectures

• Ji Lin: o3/o4-mini, reasoning stack

• Joel Pobar: inference (Anthropic), ML infrastructure

• Jack Rae: Gemini pre-training/reasoning, Gopher/Chinchilla

• Hongyu Ren: GPT-4o, post-training

• Johan Schalkwyk: speech systems, Sesame, Maya

• Pei Sun: post-training/coding for Gemini, Waymo perception

• Jiahui Yu: GPT-4o/4.1, multimodal perception

• Shengjia Zhao: ChatGPT/mini models, synthetic data

Meta essentially bought the humans who built mostly every breakthrough AI feature you've used in the past year.

What it means: Meta stopped playing defense and went full offense after a lackluster Llama 4.

They're not trying to catch up anymore. They're trying to leapfrog everyone by hiring the actual brains behind every major AI advancement.

Smart move.

When you can't out-innovate competitors, you buy their innovation teams.

Zuckerberg just turned AI development into a talent acquisition war he's determined to win.