1. Taobao Live’s digital-avatar agent adopts “harness-aware” training; team reports higher GMV and item-page views in online tests Taobao Live’s digital-avatar streamers must answer product questions, interact with viewers and carry out marketing strategies in real time.

2. OpenExecutive Open-Sources Virtual Executive System: Eight Specialist Agents Share Company Data, Decision Memory and Task Schedules OpenExecutive, a new open-source system for business advice, presents users with one virtual executive while routing their questions among eight specialist AI agents.

3. PAWBench Tests 11 Video-Generation Systems: No Model Consistently Reproduces the Probability Distribution of Physical Behaviors Generating one physically plausible video does not show that a model understands the range of ways an event could unfold.

In Brief

• Amazon adds 2 million Nvidia GPUs to AWS expansion Amazon will deploy another 2 million Nvidia Blackwell Ultra, Rubin, and Rubin Ultra GPUs in AWS data centers during 2027 and 2028, following an earlier commitment for more than 1 million GPUs. The expanded partnership also covers Nvidia networking, models, robotics technology, and server CPUs.

• Judge overturns Pentagon’s Anthropic blacklist US District Judge Rita Lin ruled that the Pentagon’s designation of Anthropic as a supply-chain risk was unconstitutional retaliation and “arbitrary and capricious.” The dispute followed Anthropic’s refusal to permit its AI systems to be used for mass surveillance of Americans or lethal autonomous weapons.

• OpenAI brings ChatGPT ads to India OpenAI will begin showing ads on ChatGPT’s Free and Go tiers in India, initially working with 50 brands through agency partners WPP and Omnicom. The company also plans to launch a self-service ad manager next month with a minimum daily campaign budget of ₹725.

• EPA proposal would shift public-notice decisions for some air permits to states The US Environmental Protection Agency has proposed eliminating the federal public-participation requirement for permits covering certain “minor” pollution sources, including some data centers and power facilities. The comment period has closed, and the agency says it is reviewing more than 4,900 submissions before finalizing the rule.

• Lambda borrows $1 billion to supply Nvidia chips to Microsoft AI cloud provider Lambda raised $1 billion in short-term private debt arranged by JPMorgan Chase to purchase Nvidia GPUs that it plans to lease to Microsoft, Bloomberg reported. The financing follows other large secured loans as Lambda expands customer-specific GPU infrastructure.

• Anthropic tests automated researchers for alignment training An Anthropic paper reports that automated research systems improved model performance across 10 benchmarks for misaligned behavior without reducing overall performance. The system searched literature, proposed and tested training methods, and retained successful approaches, though its effectiveness depended on benchmarks accurately representing the intended alignment goals.

• OpenAI hires Meta’s regional leader for Asia-Pacific expansion Sandhya Devanathan, Meta’s vice president for India and Southeast Asia, is joining OpenAI to oversee growth, enterprise adoption, partnerships, regulation, and operations across Southeast Asia and Australia. She will be based in Singapore and report to OpenAI’s Asia-Pacific managing director.

• Google hires AI researcher Barret Zoph as research vice president Barret Zoph, a Thinking Machines co-founder who subsequently returned briefly to OpenAI, has joined Google as vice president of research. Google said he will contribute reinforcement-learning and post-training expertise to Gemini.

• Instinct reaches $2.5 billion valuation while still in private beta Instinct, a personal AI assistant from Spear Street Technology that connects with users’ apps and devices, raised a $250 million Series B co-led by Index Ventures and Benchmark. The round brought its total funding to $350 million, while early access has also prompted concerns about the product’s permissions and terms of use.

• OpenAI expands its operations in Brazil OpenAI announced that it is expanding its presence in Brazil and deepening its engagement with developers, businesses, and communities to support AI adoption.

• OpenAI and Thailand launch accelerator for 10 AI startups OpenAI and Thailand’s Ministry of Higher Education, Science, Research and Innovation launched an eight-week accelerator for 10 health, wellness, and education startups. The program is intended to help participants turn AI prototypes into products.

Read the full edition →