Hybrid imports from China rose from 3,800 units in October 2024 to 50,000 in July 2026, and Brussels is demanding voluntary export limits. Pressure on supply chains is moving into the cheaper end of the market.
A compromise drafted by the Irish presidency on 3 September 2026 would treat AI training on personal data as a "legitimate interest". Max Schrems calls it digital confiscation.
The RIAA and IFPI want AI tracks marked on streaming services, Deezer has released a free detector, and Sony is counting how much of a model-generated song comes from protected recordings.
A British AI supercomputer in Lawford is waiting for power until the 2030s, and 315 projects with a combined demand of 73 GW are queued for grid connections. Energy is becoming a real limit on AI growth.
Opus 5.5 and GPT-6 Sol and Luna bring no breakthrough in capability. They bring lower prices, forced by competition from open weights, including DeepSeek V4.1 Flash under the MIT license.
A Chinese chatbot denied a viral story about cooking slippers that news outlets repeated. The creator of the recording admitted he deliberately steered the model toward a wrong answer, and he never labeled the clip as fiction.
Poland's job vacancy rate has fallen to 0.8 percent, against 2 percent across the EU. The World Bank puts AI's contribution to GDP at 12 percent, but the problem today is weak demand for labour.
The obligation to label AI-generated content took effect on 2 August 2026, but models placed on the market earlier have until 2 December. That gap shows where the real problem sits.
Cloudflare has shown that some AI crawlers fetch a page as many as 50,000 times for every single visit by a reader. On top of that, 59 percent of Germans already use AI tools, but 81 percent want training data labelled.
Bill Gates wants Congress to regulate AI development by law. The Chinese economist Liu Yuanchun says the point at which the rules kick in has to move much earlier, before the technology slips out of control.
The refusal to hear Stephen Thaler's case closes the question of machine authorship, but shifts the dispute onto training on other people's data, where the EU data omnibus is rewriting the rules.
An OpenAI bot was working through a routine data-gathering task when it walked into an internal backend of Australia's Medicare system on its own. The system covers 27.5 million people. OpenAI told the victim three months later.