The EU's first AI Act inquiry + OpenAI's power bet

Henna Virkkunen, the European Commission's Executive Vice-President for technology, said on August 29 that the AI Office has formally sent requests for information to providers of general-purpose AI models based in several regions. Euractiv reported the recipients include OpenAI, Anthropic and Google, and the demands cover model security, independent external evaluations, and monitoring of models after they reach the market. This is the first enforcement step since the general-purpose obligations became binding on August 2, and a reply the Commission judges incomplete or misleading can cost 15 million euros or 3 percent of global turnover.

Draft offering documents reviewed by the Wall Street Journal show that SB Energy, the SoftBank-controlled power developer, granted OpenAI warrants valued at $3.6 billion when they were awarded in January — a position worth $5.5 billion by the end of June. What that bought was OpenAI as anchor tenant of the PORTS-Pike campus in Ohio, the site Nvidia agreed to backstop in mid-August. SB Energy is preparing an offering of $5 billion to $7 billion as soon as September, and reports a contracted backlog above $400 billion despite operating no finished data centers yet.

Elon Musk confirmed that SpaceX is adding a blades-and-vanes foundry at its Bastrop, Texas site, where the company bought roughly 830 acres between March and June. Musk says casting those parts is the limiting factor in gas turbine production, and that doing it in-house could pull as much as 18 months out of the wait to bring new turbines online. SpaceX is targeting about 10 gigawatts of AI computing capacity by the end of 2027 — and TechCrunch notes the same turbines carry a measurable local air-quality cost, a fight xAI is already having in Memphis.

Anthropic has begun emailing users whose Claude sessions were replayed by an attacker using commodity infostealer malware, including families such as Vidar, RedLine and Atomic Stealer. Because that malware lifts an already-authenticated session cookie rather than a password, it walks past two-factor authentication and single sign-on entirely. Anthropic is signing affected accounts out, removing saved payment methods and refunding charges it judges unauthorized, and says the tell is a usage limit that refills and then drains while you are not using Claude.

Prosecutors in Taoyuan searched Unimicron's headquarters and a manufacturing plant on August 28 and questioned 14 executives and staff, alleging the company shipped circuit boards made in mainland China back to Taiwan and relabeled them as Taiwanese. A department general manager was released on bail of 15 million New Taiwan dollars, about 474,000 US dollars. This is a separate matter from the server-smuggling charges filed in Taiwan a week earlier, and it points at a different weak spot — not where AI hardware ends up, but whether the paperwork saying where it was made can be trusted.

The OpenClaw Foundation released version 2.0 of its open-source agent platform, describing it as the largest update in the project's history and one that touches installation, messaging, memory, skills, models, automations, the browser and native apps, plugins and security. It folded in more than 16,000 pull requests — roughly half of everything ever merged — from 933 contributors, 569 of them first-time. The headline changes are a rebuilt browser app and shared cloud sessions that let a team work inside one agent session together.

Caterpillar is applying what it learned automating mine sites to a company-wide AI push, chief technology officer Jaime Mineart told TechCrunch. The company runs 1.6 million connected machines generating 16 petabytes of structured data, and is spending $100 million over five years to retrain 118,000 employees whose work shifts from operating one machine to supervising several remotely. Its power-generation division, which sells into data centers, grew sales 72 percent in the second quarter of 2026.