AI Industry Development & Policy Dynamics

According to mpost.io, Meta launched Muse Spark AI on May 14, 2026, a new multimodal system enabling voice conversations with real-time visual interaction, image generation, and seamless language switching. The assistant can interrupt and switch topics mid-conversation while providing contextual

According to Beating, AI evaluation firm Vals AI released its second-generation Finance Agent v2 benchmark on May 14, testing financial analysis workflows through 927 expert-reviewed questions. GPT-5.5 topped the rankings with a 51.76% accuracy rate, closely followed by Claude Opus 4.7 (51.51%)

According to Beating, LangChain announced two major upgrades at its Interrupt conference: SmithDB, a new database built on object storage, and LangSmith Engine, an automated debugging tool. SmithDB improves query performance up to 15 times by replacing traditional local disk architecture. LangSmith

A woman in California, the United States, has filed a class-action lawsuit against OpenAI, claiming violations of privacy laws. She alleges that ChatGPT, without users’ consent, sends users’ query contents to Meta and Google through tracking tools. The case is seen as a symbolic first federal court challenge directly targeting how AI chatbots handle data, drawing intense attention to AI privacy protection. （ChatGPT involved in a teen suicide case! OpenAI: has strengthened GPT-5’s psychological p

According to monitoring by Beating, three technical leads directly responsible for Grok model development departed xAI within the past week. Tianyi Zhang, who led search and factuality post-training, confirmed his departure today; Juntang Zhuang, pre-training team lead, left earlier this month; and

According to Beating, Tencent Cloud recently open-sourced TencentDB Agent Memory, a local-first memory engine for AI agents. The system reduces token consumption by 61% in complex workflows—decreasing usage from 221.31M to 85.64M in WideSearch tasks—while improving task completion rates by 51.52%. T

Foxconn expects strong demand for AI servers, with second-quarter AI cabinet shipments forecast to grow in double-digits compared to the first quarter. The company projects full-year shipments to grow more than 100 percent year-over-year.

Tianyi Zhang, who led xAI's search and factuality post-training team, announced his departure from the company today (May 14). Over the past year, Zhang oversaw the development of Grok's real-time search and Agent capabilities, serving as the core architect behind the system's ability to retrieve ac

Key Insights Covenant AI exited Bittensor and sold over 37,000 TAO tokens, triggering heavy selling pressure and wiping out recent market gains. Sam Dare accused Jacob Steeves of centralized control, fueling concerns among traders about governance transparency within the broader Bittensor ec

According to Ramp AI Index, Anthropic's enterprise adoption rate reached 34.4% in April, surpassing OpenAI's 32.3%, marking the first time the AI company has led in corporate adoption. Anthropic's rate rose 3.8% month-over-month, while OpenAI declined 2.9%. Over the past year, Anthropic's enterprise

According to Reuters' exclusive report today, approximately 10 Chinese companies including Alibaba, Tencent, ByteDance, and JD.com have been approved to purchase Nvidia's H200 chips, with each customer limited to 75,000 units. Lenovo and Foxconn were approved as distributors. However, no chips have

According to Abacus.AI CEO Bindu Reddy, Google plans to unveil Gemini 3.2 Flash during its I/O conference on May 20, with performance reaching 92% of GPT-5.5 on coding and reasoning tasks while cutting inference costs to just one-fifteenth to one-twentieth of the latter. Most queries will have

At Think 2026 conference on May 5, IBM unveiled expanded capabilities for its Confluent platform to support real-time data processing for AI-driven operations. The company introduced tools aimed at helping enterprises improve AI deployment, governance, and operational scalability. According to

On May 14, a social media post by New York Post reporter Emily Goodin said that the White House press corps saw Elon Musk in Beijing and conducted interviews. After the summit, Musk said the talks went “very smoothly,” adding that “a lot of good things are happening.” Jensen Huang said the talks went “very smoothly,” adding that both heads of state of China and the U.S. were “absolutely fantastic.” After making a “yeah” hand gesture at the camera, Cook raised his thumbs-up. Confirmation reaction

Two major updates appeared around May 14 for Doubao, the AI assistant under ByteDance: leaked internal-test screenshots show that the “Scan to Pay” feature now has a “Pay Orders” entry at the bottom; the App Store page was updated in parallel to confirm that Doubao will roll out three tiers of paid subscription plans on top of the free version, with monthly fees of 68 Yuan, 200 Yuan, and 500 Yuan, respectively. Scan to Pay feature: confirmed status and unconfirmed parts in internal-test screensh

CME Group, the world’s leading derivatives market, announced that it will partner with Silicon Data, a leader in GPU market intelligence, to launch the world’s first “Compute Futures” market later this year. The move aims to address the rising price volatility and growing risk management needs in the computing market, whose value is measured in the multi-trillions of Dollars. The contract will be based on Silicon Data’s daily GPU benchmarks for on-demand rental rates, providing the market with a

On May 14, OpenAI released a statement regarding the "Mini Shai-Hulud" supply chain attack targeting the popular open-source library TanStack and multiple npm packages, confirming that its security team found no evidence of user data breach or unauthorized access following an internal system

According to Ads3's official announcement on May 13, Funton.AI and Ads3 have entered into a strategic partnership to integrate Ads3's AI-powered advertising infrastructure into Funton.AI's Web3 gaming platform. The collaboration aims to improve user acquisition and visibility for the gaming

According to a May 13 announcement on the WhatsApp official website, the social messaging app has launched a “Meta AI Private Processing” feature called “Meta AI Private Processing” for private conversations. It uses six core technologies, including trusted execution environment TEE and RA-TLS end-to-end encryption, to ensure that user messages are processed in a secure environment, and even Meta cannot access the conversation content. After message processing, the messages automatically disappe

According to InitVerse's official announcement on May 13, 2026, InitVerse and UXLINK announced a strategic partnership to accelerate decentralized application development and expand social onboarding across Web3. The collaboration combines InitVerse's AI-powered development framework with UXLINK's s