AI News

Follow the latest AI and automation news

28 Jul 202614 Safar 1448 AH

Ookla unveils trends in AI cloud latency

Recent research by Ookla revealed that network responsiveness for AI applications varies significantly by location. The study found that fibre infrastructure plays a bigger role than population size or local data centers. For instance, the fastest median latency recorded was 104 milliseconds in Providence, Rhode Island, while Roswell, New Mexico, had a latency of 146 milliseconds.

Business

28 Jul 202614 Safar 1448 AH

Samsung and Broadcom sign an AI chip deal

Samsung Electronics has signed an agreement with Broadcom valued at over $200 billion to enhance collaboration on advanced chip technologies through 2030. The partnership aims to deliver integrated semiconductor solutions for the AI era, focusing on high bandwidth memory (HBM) for Broadcom's next-generation AI accelerators. This collaboration will also leverage Samsung's advanced process technologies for foundry solutions. This strategic alliance marks a significant step for Samsung to strengthen its position in the chip market, aiming to close the gap with foundry leader TSMC and meet the growing demand for AI technologies.

Governance

28 Jul 202614 Safar 1448 AH

Proofpoint introduces Active Exploits Protection

Proofpoint, a leader in cybersecurity, has launched a new solution called Active Exploits Protection aimed at empowering organizations to defend against the rising threats posed by AI. This solution identifies vulnerabilities actively exploited in the wild, translating that intelligence for immediate protection.

💡

Deconstructing the Kimi K3 Technical Report: Moonshot AI Starts Setting Problems for DeepSeek as Two Scaling Axes Redefine the Frontier

What challenges do open-source models face in pre-deployment compute?

Agents

Proofpoint joins the OpenAI Daybreak Cyber Partner Program

28 Jul 202614 Safar 1448 AH

Agents

28 Jul 202614 Safar 1448 AH

Proofpoint joins the OpenAI Daybreak Cyber Partner Program

Proofpoint, a leading cybersecurity company, has announced its selection to join OpenAI's Daybreak Cyber Partner Program. This initiative aims to assist trusted cybersecurity organizations in integrating advanced AI capabilities into defensive security operations. As cyber threats evolve, security teams are under increasing pressure to investigate incidents quickly and analyze larger data volumes.

Models

China Surpasses US With Two-Thirds of Global LLM Calls as Xiaomi MiMo-V2.5 Ranks First in Weekly and Monthly Token Usage Worldwide

28 Jul 202614 Safar 1448 AH

Models

28 Jul 202614 Safar 1448 AH

China Surpasses US With Two-Thirds of Global LLM Calls as Xiaomi MiMo-V2.5 Ranks First in Weekly and Monthly Token Usage Worldwide

Chinese AI models now account for two-thirds of global inference calls, with Xiaomi's MiMo-V2.5 leading at 10.5 trillion tokens weekly. As of July 26, Chinese models hold a 63.5% market share compared to 35.5% for U.S. models, indicating a significant market shift. The rapid improvement in Chinese model quality and aggressive pricing strategies have made these models more accessible to global developers. Since the public beta launch of MiMo-V2.5 on April 23, 2026, its weekly token volume surged from 1.46 trillion in May to 10.46 trillion in July, marking a 616% increase in just two months.

Arab AI Scene

Abu Dhabi to launch world’s first AI Judicial Platform

28 Jul 202614 Safar 1448 AH

Arab AI Scene

28 Jul 202614 Safar 1448 AH

Abu Dhabi to launch world’s first AI Judicial Platform

The Abu Dhabi Judicial Department has announced the implementation of its AI Judicial Platform, touted as the world's first integrated system of its kind. The platform will support legal decision-making under full human supervision, with the first phase set to begin in September 2026 and a phased rollout planned over 18 months.

Models

Kimi K3 Officially Open-Sourced: Moonshot AI Releases 2.8 Trillion Parameter Model Weights, Technical Report, and Three Infrastructure Technologies

28 Jul 202614 Safar 1448 AH

Models

28 Jul 202614 Safar 1448 AH

Kimi K3 Officially Open-Sourced: Moonshot AI Releases 2.8 Trillion Parameter Model Weights, Technical Report, and Three Infrastructure Technologies

Moonshot AI released the open-source Kimi K3 on July 27, featuring 2.8 trillion parameters and 896 routed experts. Compared to the previous K2.5 generation, K3's parameter scale increased approximately threefold, while computational efficiency improved by 2.5 times through new technologies. The architectural innovations include a hybrid attention structure that combines linear attention with gated multi-head latent attention, enhancing long-context processing. Additionally, the Stable LatentMoE routing mechanism addresses the expert collapse issue, stabilizing router gradient dynamics.

Governance

Microsoft unveils AI security tools it says outperform competing platforms

27 Jul 202613 Safar 1448 AH

Governance

27 Jul 202613 Safar 1448 AH

Microsoft unveils AI security tools it says outperform competing platforms

Microsoft has announced new tools aimed at helping customers streamline and automate the identification and reduction of their cybersecurity risks. This announcement comes less than a week after OpenAI lost control of two security models, leading to a breach of Hugging Face's servers. The breach involved a "swarm of tens of thousands of automated actions" that stole Hugging Face's internal credentials.

Arab AI Scene

UAE's G42 joins Open Secure AI Alliance started by Nvidia

27 Jul 202613 Safar 1448 AH

Arab AI Scene

27 Jul 202613 Safar 1448 AH

UAE's G42 joins Open Secure AI Alliance started by Nvidia

On Monday, Abu Dhabi-based AI company G42 announced its membership in the Open Secure AI Alliance led by Nvidia. This alliance includes prominent firms endorsing open AI models that can be freely downloaded, shared, and modified. Peng Xiao, CEO of G42, emphasized that the future of AI will be shaped not only by model capabilities but also by how securely nations and enterprises can build on them. He highlighted the importance of open models in fostering security, trust, and sovereign control from the outset.