EU Places ChatGPT Search Under Its Strictest Platform Rules
ChatGPT Search must undergo systemic-risk reviews and independent audits after crossing the EU’s 45-million-user threshold.
August 30, 2026
All briefings and quick hits published that day.
- ⚡ Quick hits
- AIGC Hailuo demonstrated H3 Max generating interactive game interfaces at near-real-time speed from user actions. X / Hailuo AI ↗
- ⚡ Quick hits
- Agents Lovart introduced a workflow for creating high-fashion brand campaigns with substantially smaller production budgets. X / Lovart ↗
- AIGC Pocket can turn game concepts into playable prototypes, but Meta now controls the generated results and platform. Ars Technica ↗
- Agents OpenAI and rival laboratories are reportedly buying tens of thousands of Mac minis to train computer-use agents. The Decoder ↗
- Industry Fourth Paradigm signed a computing partnership with Huawei and joined the first adopters of its highest-end domestic AI infrastructure. 量子位 ↗
DiDi Starts Driverless Passenger Trials With R2 Robotaxi
Passengers in parts of Beijing and Guangzhou can now summon DiDi’s purpose-built R2, equipped with 33 sensors and redundant controls.
- ⚡ Quick hits
- AIGC Wan shared a commercial director’s workflow for producing a 30-second advertisement with Wan 3.0. X / Wan ↗
- Models MiniMax made its H3 and H3 Max models available for testing through Vercel. X / MiniMax ↗
- AIGC Wan highlighted an AI film whose character remained visually consistent across multiple scenes generated with Wan 3.0. X / Wan ↗
- Industry Lingxi Zhiyong positioned its ROSS Harness among China’s leading industrial embodied-AI systems after national evaluations. 量子位 ↗
- Agents AQuA introduced an evolving quantitative-research agent designed to produce more auditable backtesting results. 量子位 ↗
- Models Apodex 1.1 scored 44 on Artificial Analysis’s Intelligence Index and reached 1348 Elo on GDPval-AA v2. X / Artificial Analysis ↗
- Policy U.S. barriers targeting Chinese drones and robots may be offset by China’s manufacturing scale, TechCrunch reported. TechCrunch ↗
OpenClaw 2.0 Turns Personal Agents Into a Shared Platform
The open-source agent adds remote sessions, interactive dashboards and scoped secret handling in a major architectural overhaul.
- ⚡ Quick hits
- Agents Simon Willison published a technical walkthrough examining how the new ChatGPT Work environment operates. Simon Willison ↗
- Research The Kuleshov Group published a practical guide to building diffusion language models using masking, refinement, and variable-length generation. Kuleshov Group ↗
- Industry Microsoft reportedly tightened employee AI budgets after one worker consumed $28,000 in tokens within 28 days. IT之家 ↗
- Agents A TTFT-focused benchmark compared inference APIs to identify the lowest-latency options for voice and real-time AI agents. MarkTechPost ↗
- Models Together AI reported GLM-5.3 had a substantially lower hallucination rate than Claude Fable 5 and GPT-5.6 Luna in its evaluation. Together AI on X ↗
- Agents Google introduced EnvHarness, a programmable layer that converts static agent environments into adaptive worlds for training and evaluation. MarkTechPost ↗
- Agents GitHub said automated Dependabot pull requests comprised roughly one-sixth of commits on Microsoft’s GCToolkit, illustrating alert-fatigue risks. GitHub on X ↗
- Industry xAI’s faster route to installing gas turbines for AI infrastructure could increase local air pollution, according to TechCrunch. TechCrunch ↗
- Industry Databricks outlined a new batch of platform features and explained how they fit into its broader data and AI architecture. Databricks on X ↗
- Policy Texas Governor Greg Abbott blocked state funding for additional Flock automated license-plate cameras amid surveillance concerns. The Verge ↗
- Industry Caterpillar is applying operational lessons from autonomous mining equipment to industrial AI deployment and management. TechCrunch ↗
- Industry An industry analysis argued that young AI startups commonly postpone infrastructure optimization until scale makes inefficiency financially painful. SiliconANGLE ↗
- Safety SiliconANGLE detailed four safeguards intended to reduce harmful or unauthorized behavior by autonomous AI agents. SiliconANGLE ↗
- Industry Employee-review analysis found workplace sentiment toward corporate AI adoption becoming increasingly negative amid pressure, disruption and uncertain benefits. The Decoder ↗
- Industry Meta is developing robots to inspect, maintain and operate data centers as AI infrastructure expands. Ars Technica ↗