It’s Not Me, It’s Your Transparency Law.

Meta is pulling the plug on all political, electoral, and social issue ads across the European Union starting in October 2025. Why? Because the EU’s new rules, meant to make political ads more transparent, are being called “unworkable” by the tech giant. The policy shift will affect Facebook, Instagram, and Threads in all 27 EU member states. Meta says complying with the EU’s Transparency and Targeting of Political Advertising (TTPA) regulation would be too complicated and legally murky. So rather than risk fines or legal gray zones, they’re walking away from the whole thing.

The TTPA law officially took effect in April 2024 but kicks into full gear on October 10, 2025. It requires platforms to label political ads, show who paid for them, and link them to specific campaigns or legislative processes. The ads also have to be archived in a public database and can only be targeted under strict conditions. That may sound reasonable if you’re thinking about democracy and transparency, but for companies like Meta, it means building entire new compliance systems. Meta says these requirements introduce a level of “complexity and legal uncertainty” that’s just not worth the hassle.

Meta isn’t alone. Google already announced in November 2024 that it would stop accepting political and issue-based ads in the EU too. So what we’re seeing isn’t just one company crying foul, it’s a broader industry pushback against the EU’s efforts to rein in Big Tech’s power in politics. The regulation includes steep penalties for noncompliance, with fines reaching up to 6% of a company’s annual global turnover. For companies with billions in revenue, that’s a serious threat.

It’s worth noting that Meta’s decision doesn’t silence politicians or end political debate on its platforms. Users can still post and share political content freely. Politicians can still speak. They just can’t use Meta’s advertising tools to boost that content to a wider audience. In other words, Meta is cutting off the paid megaphone, not the microphone. The implications are big. Political campaigns in the EU will now lose one of their most powerful digital tools for reaching voters, particularly during high-stakes election seasons.

This move marks a major flashpoint in the ongoing tug-of-war between Silicon Valley and Brussels. The EU wants more transparency and accountability in online political discourse. Big Tech says the rules are too burdensome and unclear. And in the end, voters in Europe might be the ones caught in the middle, left with less clarity about who’s trying to influence them and fewer tools for civic engagement. Whether this results in better or worse democratic outcomes is still up for debate. But one thing’s clear: the way politics works online in Europe is about to change.

Rapid Fire

👾 OpenAI is gearing up to launch GPT-5 in early August 2025, and this one isn’t just another model upgrade, it’s a full system reboot. According to The Verge and other outlets, GPT-5 will unify the GPT and o-series tech under one roof, combining the raw power of GPT with the o3 model’s advanced reasoning skills. Think smoother logic, sharper problem-solving, and fewer moments where the AI fumbles a basic task. CEO Sam Altman even teased its capabilities on Theo Von’s podcast, calling its performance "weird", in a good way. If that’s not classic Altman-speak for "this thing is wild," what is?

But it’s not just one model. GPT-5 is launching in multiple flavors. The mini version will be available through ChatGPT and API, while a leaner nano variant will be API-only. Microsoft has been busy since May beefing up servers, prepping for the expected surge in use. Meanwhile, OpenAI’s also set to release its first open-weight model since GPT-2, this one aimed at the reasoning crowd, dropping before the end of July. All of it points to one thing: OpenAI is making a hard play to dominate both casual and power users, just as competition from Google and Anthropic heats up.

👨‍💻 Google just entered the “vibe-coding” game with Opal, a new tool that lets anyone spin up AI-powered mini web apps using plain English and drag-and-drop workflows. No code, no config—just vibes. Users describe what they want, and Opal builds a working app by chaining together AI models and tools behind the scenes. Then you can tweak, remix, and share your creation straight from your browser. Think Canva meets ChatGPT, but for app building.

This move isn’t just about showing off cool tech. It’s Google’s play to democratize software development and stay ahead in the AI arms race. While devs already use tools like Replit and Figma for no-code prototyping, Opal targets casual creators and power users alike with its visual editor and “vibe-first” approach. It's available now in the U.S. through Google Labs, and if it catches on, it could turn prompt-writing into the next big creative skill, right alongside graphic design and video editing.

📰 AI isn’t just generating poems and weird pizza recipes anymore. It’s now moving serious traffic. In June 2025, AI platforms pushed over 1.13 billion referrals to the top 1,000 websites. That’s a 357% jump from last year. The lion’s share of that traffic came from ChatGPT, which sent over 80% of those referrals. But let’s not get ahead of ourselves. Google Search still towers over the competition, casually throwing out 191 billion referrals in the same timeframe. So while AI is sprinting, Google is still lapping everyone on the track.

What’s wild is where that AI traffic is going. News and media outlets like Yahoo, Reuters, and The Guardian raked in millions of visits from AI-driven links. This matters because it hints at how we’re shifting the way we get information. More than 120 million Americans are expected to use generative AI by the end of 2025. But here’s the kicker. Traditional search traffic is actually down, dropping by 6.7% year-over-year. So yeah, AI isn’t replacing Google yet, but it’s starting to nibble at the edges.

Tech Radar

Despite U.S. export bans, Nvidia’s top-tier AI chips are alive and well in China, just not through official channels. A booming black market repair industry has sprung up in Shenzhen, with boutique shops fixing hundreds of smuggled H100 and A100 GPUs each month. Some of these chips have been running nonstop for years, grinding through data and now breaking down without access to legal repair services from Nvidia. With no help coming from the source, Chinese firms have turned to their own underground tech support scene, and business is booming.

The repair surge isn’t just about wear and tear. It’s proof that smuggling is happening at scale. Intelligence estimates say over $1 billion worth of banned chips slipped into China in just three months. These chips don’t sneak in through back alleys, they come through Hong Kong or Southeast Asia, disguised in commercial shipments. Even government tenders show military-linked buyers snapping up restricted GPUs. And while the U.S. tries to plug the leaks with proposed tracking laws, China’s appetite is already shifting to Nvidia’s even newer B200 chips, despite each server costing north of $400,000. Bottom line: export bans may slow the flow, but they haven’t stopped the demand.

Recently Deployed

Tesla’s robotaxi dreams are inching into San Francisco, but not without some caveats. After kicking things off last month in Austin, Tesla is now launching a more limited version of the service in the Bay Area, possibly as soon as this weekend. But unlike Austin, where the driver’s seat was empty and a Tesla employee rode shotgun, San Francisco’s version will feature someone actually behind the wheel. Why? Because Tesla doesn’t have the California permits to go full robo yet. So while this is technically a robotaxi, it’s got training wheels.

There’s also the question of legality. As of now, Tesla hasn’t secured the permits from the California Public Utilities Commission needed to pick up paying customers in fully autonomous vehicles. For now, rides are restricted to Tesla employees or invited users. The service will be geofenced but still cover a big chunk of the Bay, think Marin to San Jose. It’s a cautious rollout, especially compared to Waymo, which already has fully driverless cars cruising SF and pulling in real revenue. Still, with Elon Musk calling California regulators “reasonable,” Tesla seems confident it’ll get there. Just... not this weekend.

That’s a wrap. 

EU elections are going manual, GPT is going “weird,” and Elon’s robotaxis are still waiting for permission to drive themselves. Welcome to the future, where everything’s powered by AI, but still needs a human chaperone.