AI Briefing

Auto China 2026 just gave us the clearest signal yet that commercial L3 driving is no longer a 'someday' story, while Google quietly planted a $15 billion flag in India that could reshape where the world's AI actually gets built. Meanwhile, Meta's buying robot brains, Stanford says human scientists still run circles around AI agents, and the EU's AI Act deadline drama just got messier.

01ProductCGTN ↗

Huawei Launches Commercial L3 Autonomous Driving at Auto China 2026, Legally Letting Drivers Go Hands-Free

Huawei's ADS 5.0 became the first system in China to offer license-plate-ready Level 3 autonomous driving for highways, meaning drivers can legally take their hands off the wheel at speeds of 60-120 km/h. The system uses an 896-line LiDAR and a self-developed vehicle OS, with Huawei revealing its cloud training compute has exploded 21-fold to 60 EFLOPS in 29 months. By year-end, over 80 car models from partners including BYD, Audi, and Mercedes-Benz will ship with the system, targeting 3 million cumulative installations.

02IndustryGoogle Blog ↗

Google Breaks Ground on $15 Billion AI Hub in India, the Largest Single Foreign Investment in Indian History

Google officially broke ground in Visakhapatnam on a five-year, $15 billion AI infrastructure project that will include India's first gigawatt-scale AI hub across three data center campuses, a new international subsea cable gateway, and partnerships with AdaniConnex and Airtel. The investment dwarfs any previous single foreign direct investment in India and is expected to generate at least 200,000 direct and indirect jobs. The move signals that the AI infrastructure race is no longer just a US-China story.

03IndustryTechCrunch ↗

Meta Acquires Assured Robot Intelligence to Power Its Humanoid AI Ambitions

Meta acquired Assured Robot Intelligence (ARI), a startup building foundation models for humanoid robots that can understand, predict, and adapt to human behaviors. ARI's co-founders Lerrel Pinto (NYU professor, former OpenAI researcher) and Xiaolong Wang (UC San Diego, former Nvidia researcher) will join Meta's Superintelligence Labs. The deal puts Meta squarely in the humanoid robotics race alongside Tesla, Figure, and 1X, targeting a market analysts project could reach $5 trillion.

04ResearchNature ↗

Stanford AI Index 2026: Human Scientists Still Trounce AI Agents on Complex Research Tasks

The Stanford Institute for Human-Centered AI's annual index report found that despite a 30-fold explosion in AI-mentioning science papers since 2010, there's scant evidence that AI agents are actually improving research productivity. AI agents still badly underperform human scientists on complex multi-step tasks, and Princeton researcher Arvind Narayanan warned the boom is 'happening too fast, without giving scientific norms time to adjust.' The report lands as 96% of enterprises claim to be running AI agents in production.

05PolicyIAPP ↗

EU AI Act Reform Talks Collapse After 12-Hour Marathon, August 2026 High-Risk Deadline Holds

After 12 hours of negotiations that began April 28 and ran into the early morning, the European Parliament and Council of the EU failed to agree on reforms to the AI Act's high-risk rules under the Digital Omnibus proposal. The collapse means the original August 2, 2026 compliance deadline for high-risk AI systems remains legally binding, with no guaranteed delay. Companies deploying AI in healthcare, hiring, law enforcement, and critical infrastructure now face a hard three-month countdown with no certainty the rules will be softened.