AI Dispatch | India Expands AI Testing Push
Here are the top tech stories from our weekly AI news wrap-up (31 July-7 August).
Topics
News
- Temasek, Seraphim Back Pixxel in $100 Million Round
- UN Rights Chief Warns Advanced AI Could Threaten Humanity
- Tata-owned JLR to Cut About 4,000 Bobs Amid Tariff, China Pressure
- Anthropic Walks Away From $6 Billion Decart Deal
- Hyundai India Targets 20% Women in Executive Workforce by 2030
- Security Breach Drains $320 Million in Bitcoin From Liquid Network
[Image: Nomita Samaiyar/MITSMR India]
India strengthened its AI evaluation push this week as AI4Bharat and Josh Talks widened Voice of India into a broader platform for testing models and deployed applications under local conditions. The government also called for human oversight and greater technological autonomy, while Anthropic prepared to process Claude requests in the country.
Globally, Meta and Anthropic disclosed testing failures that allowed AI models to interact with live systems, Alphabet reshuffled its AI leadership, Alibaba launched a 2.4-trillion-parameter Qwen model, and Google paused an image-generation feature in Google Earth over misuse concerns.
Voice of India Broadens AI Evaluation
AI4Bharat and Josh Talks expanded Voice of India beyond speech recognition to evaluate language models, voice systems, document processing, safety and deployed applications. The platform will test systems under Indian linguistic, cultural and operating conditions.
Read more: AI4Bharat, Josh Talks Launch India AI Evaluation Platform
India Calls for Human Oversight
IT Secretary S. Krishnan said organizations should retain human responsibility for AI output and avoid excessive dependence on a single foreign supplier or country. He also said India would eventually need AI regulation but provided no timetable.
Read more: IT Secretary Krishnan Urges Human Oversight as AI Adoption Grows
Anthropic Adds India Processing
Anthropic plans to process Claude requests in India through Amazon Bedrock in the coming weeks. The move could help regulated companies address data-location requirements, although model availability, pricing and launch timing remain unclear.
Read more: Anthropic Brings Claude Processing to India
AI Tests Reach Live Systems
Meta and Anthropic disclosed separate incidents in which models accessed or exploited live systems after testing environments were mistakenly connected to the internet. The cases, which follow similar failures involving OpenAI and Hugging Face, point to weak operational controls rather than models escaping containment.
Read more: Anthropic Says Claude Models Breached Three Companies During Cyber Tests and Meta AI Model Exploits Third-Party System After Testing Error
Alphabet Reshuffles AI Leadership
Alphabet moved DeepMind chief Demis Hassabis into a research-focused role as chief scientist and chairman. Koray Kavukcuoglu will take operational control while four senior executives leave to establish Discovery Loop.
Read more: Alphabet Reshuffles AI Leadership
Alibaba Launches Largest Qwen Model
Alibaba introduced Qwen3.8-Max, a multimodal model with 2.4 trillion parameters and a context window of up to one million tokens. The company said it would release the model weights, although some performance claims remain independently unverified.
Read more: Alibaba Launches Its Most Powerful Qwen Model Yet
Google Pauses Earth AI Image Tool
Google suspended an image-generation feature in Google Earth after users placed fabricated events over real geographic imagery. The company said it would strengthen safeguards before restoring the tool.
Read more: Google Pauses Earth AI Image Tool Over Fabricated Satellite Images
Hitachi Names AI Unit Chief
Hitachi appointed Anand Birje chief executive of its digital engineering and AI business unit. The division combines GlobalLogic, Hitachi Digital Services and Hitachi Vantara capabilities across engineering, cloud, data and AI.
Read more: Hitachi Names Anand Birje AI Unit CEO
