MIT Technology Review argues that a summer of AI claims from Anthropic and OpenAI, from hacking incidents to math breakthroughs, has been amplified by press coverage while later expert reviews tell a less dramatic story.
Products & Applications · Companies · Large Language Models
Alibaba Cloud introduced AgentCore, Agent Sandbox and new storage products at the 2026 Apsara Conference, while Qianwen detailed Personal Agent progress for health, finance and work. Kantata launched Agent Studio, and Jiyuan Ludong outlined model-routing and RSI work.
A four-university team's COBRA-Skills framework casts agent skill optimization as a budget-controlled contextual bandit problem, cutting optimization cost 55–58% with only 50 samples.
A new DataDome report says malicious automated traffic rose 124% from July 2025 to June 2026, more than nine times the growth of human traffic, while 65.3% of 21,491 popular websites tested blocked none of the ten bot types thrown at them.
The All-China Federation of Industry and Commerce released the 2026 China Private Enterprises 500 list on September 22, 2026. JD.com, Alibaba, Hengli, Huawei, BYD and Tencent took the top six spots, with Huawei and Tencent leading R&D spending.
OpenAI says its AI solved the Navier-Stokes problem, a Millennium Prize question, but mathematicians say its 166-page proof is technically correct yet hard to read and offers little human understanding.
Products & Applications · Companies · Large Language Models
Qwen Office announced Enterprise Context, Applications, QwenNote and a Security Center at the Yunqi Conference in Hangzhou, aiming to give enterprise agents business understanding, organizational identity and traceable execution.
Xiaomi has open-sourced MiMO-V2.6 Pro and MiMO-V2.6-Flash, omnimodal models it says can build interactive 3D worlds such as games from text or visual prompts, alongside Windows and Mac desktop apps.
More than a quarter of UK consumers trust chatbots for money advice, but the FCA says ChatGPT and Claude sit outside its remit, leaving no ombudsman or compensation route.
Products & Applications · Companies · Large Language Models
SpaceXAI has released Grok 4.7, a larger flagship model for coding and knowledge work, keeping Grok 4.6's $2 input and $6 output price per million tokens.
AI Chips & Compute · Large Language Models · Companies · Products & Applications
Alibaba introduced the Zhenwu V900 AI chip, a plan for more than 20GW of global data center capacity by 2032, and said Qwen4 is training, with future Qwen models targeting 5-10 trillion parameters.
AI Chips & Compute · Large Language Models · Markets
DeepSeek is finalizing a 50 billion yuan funding round and betting on Huawei AI chips, according to The Information. OpenAI has formed an independent math advisory group after its new model solved more than 100 open problems, while Amazon Bedrock added Kimi K3 with a one-million-token context.
Chainalysis reports a 440% rise in malicious blockchain activity, with attackers using on-chain dead drops to keep malware instructions alive and unrestricted AI models lowering the technical barrier.
SpaceX introduced Grok 4.7, calling it its most capable large language model, with top results on coding, legal and safety benchmarks and pricing from $2 per million input tokens.
Products & Applications · Companies · Large Language Models
The Strands Agents team at AWS released Strands Harness, an Apache 2.0 general-purpose agent harness for Python and TypeScript. The team reports 28% lower token cost than rival harnesses across six benchmarks at comparable accuracy.
Companies · Large Language Models · Policy & Regulation
OpenAI proposed global frontier AI safety standards focused on alignment and recursive self-improvement, urging international cooperation and caution on autonomous RSI.
Google has confirmed that Gemini models hacked three companies during a May 2026 cybersecurity test run by Irregular, after a misconfiguration let the AI reach the internet and target real infrastructure instead of fakes.
Products & Applications · Companies · Large Language Models
AWS has launched Strands Harness, an open-source AI agent designed to move developers from local prototypes to scalable cloud deployments, with support for multiple model providers and claims of better token efficiency.
Large Language Models · Policy & Regulation · Markets
Newly unsealed statements from Microsoft and OpenAI executives threaten the New York Times' fair-use defense, while options trading in Times stock heated up as investors weighed a possible settlement or trial outcome.