Brussels will propose barring under-15s from chatbots and feeds
The AI slowdown debate went global today and every capital answered with its own agenda: Brussels reached for a ban, Berlin for a veto on slowing down, and Nvidia quietly repriced what a single desk AI box can hold.
The European Commission will unveil an EU KIDS Act on Thursday that restricts under-15s from social media, video-sharing platforms, online games and — explicitly — AI chatbots and AI companions, according to preparatory documents seen by Politico Europe and Reuters. Commission President Ursula von der Leyen and tech chief Henna Virkkunen present it Thursday, with details expected in von der Leyen's State of the Union address Wednesday; a Commission spokesperson confirmed the act is coming. The plan is not a flat prohibition but an age ladder: guardian-controlled accounts on services meeting strict safety standards below 13, introductory accounts between 13 and 15 with limits on contacting strangers and on screen time, and self-serve accounts from 15. Platforms would have to verify age at signup (games, before download), drop addictive design and harmful feeds, ship reporting and parental tools, and pay a supervisory fee that funds the regulator — with an enforcement architecture modeled on the Digital Services Act, which is already the frame under which the EU put ChatGPT in the Digital Services Act's strictest tier. Educational tools are carved out, and nothing becomes law before months of negotiation with member states and the Parliament. Our read: grouping chatbots with feeds and games is the consequential move — it treats an AI companion as a platform with an audience rather than a product with a feature list, and it makes age verification a compliance obligation on OpenAI and Anthropic, not only on Meta.
Nvidia announced the RTX Pro 5500 Blackwell Workstation Edition, a card that takes the exact GB202 die and 21,760 CUDA cores inside the GeForce RTX 5090 and gives it 84GB of ECC GDDR7 instead of 32GB. PNY's listing and Nvidia's own product page rate it at 1,398 GB/s; Tom's Hardware's analysis shows 28 GDDR7 modules clamshelled 14 to a side across a 448-bit interface, which makes Nvidia's published "416-bit" figure arithmetically impossible for 84GB, as Notebookcheck independently flagged. The card runs at 600W in a two-slot, rack-workstation form factor, with a liquid-cooled variant to follow and Multi-Instance GPU support for one 84GB instance or two isolated 42GB ones.
The catch is bandwidth: 22% below both the RTX 5090 and the flagship RTX PRO 6000, and only 4% above the RTX Pro 5000, so this is a capacity play, not a speed one. Tom's Hardware's framing is that Nvidia makes materially more margin rebinning a GB202 into a professional SKU than selling it as a gaming card, and the 5500 fills the gap between the RTX PRO 6000 (96GB, listed at $15,599) and the RTX Pro 5000 72GB ($9,209). For anyone running models locally, the practical meaning is that 84GB on one card — enough to hold a quantized 70B with a long context, or an unquantized mid-size MoE — is now an official Nvidia product tier rather than a modder's trick, the desk-side complement to the way Nvidia's free tool turns idle PCs into one local AI cluster.
Germany's government said halting frontier AI development is "not a viable option" for Europe, and Berlin is raising AI governance at G7 level. A spokesperson for Germany's digital affairs ministry tied the position to digital sovereignty — Europe needs more AI development, not less — while saying the government takes public warnings from leading developers seriously and watches closely. The ministry's specific red line was the kind of incident the industry has spent a month discussing: agents capable of breaking out of a test environment and reaching other systems, as OpenAI's did in July, would be "an entirely new threat scenario" needing a new policy response, and effective oversight requires both the United States and China at the table. Spain's digital minister Óscar López heard something different in the same debate, telling state broadcaster TVE that the voices around Europe sound like they are converging on an international agreement to limit AI risk, in language he likened to nuclear non-proliferation. Britain's prime minister's office welcomed labs talking openly about risk but said regulation should follow evidence, not speculation, and cautioned that the existing framework should not be assumed sufficient. The split is the story: Washington and Beijing both rejected the pacing ask, and Europe can't agree on whether it wants a treaty or a head of steam.
What to watch: von der Leyen's State of the Union on Wednesday, then Thursday's KIDS Act text — whether "AI companions" stays inside the ban's scope after industry lobbying.
If under-15s lose access to AI chatbots but keep it to search, homework help and voice assistants, has Brussels protected kids or just moved the risk to whoever builds the least-regulated assistant? Tell us in the comments.
Sources: Politico Europe · Reuters · RTÉ · Tom's Hardware · VideoCardz · TechPowerUp · Notebookcheck · Reuters · PYMNTS
