AI News

Follow the latest AI and automation news

24 Sept 202613 Rabiʻ II 1448 AH

Ringg’s AI agents resolve up to 65% of customer calls with OpenAI

Ringg has announced the launch of multilingual agents powered by GPT-5.6, enabling services across voice, chat, WhatsApp, and web platforms. This innovation demonstrates Ringg's ability to reduce costs by 90% compared to the previous version, GPT-4.1, significantly enhancing its operational efficiency. This move marks a major shift in how language services are delivered, allowing businesses to improve their customer communication across multiple platforms.

Agents

24 Sept 202613 Rabiʻ II 1448 AH

Securing new era of AI identities

AI adoption is rapidly increasing in the Middle East, introducing a new generation of identities into enterprise environments. In an interview with TahawulTech.com at GISEC Global 2026, Kamel Heus, Vice President of Sales-MEA at Saviynt, discussed the Zuma AI Identity Security platform. This platform focuses on managing the lifecycle of AI agents and ensuring the security of their identities.

Models

24 Sept 202613 Rabiʻ II 1448 AH

Gemini 4 is almost ready, says new Google DeepMind chief

Google is nearing the launch of its Gemini 4 model, having lagged behind competitors in flagship AI releases. Koray Kavukcuoglu, head of Google's DeepMind division, stated that the model is currently in the refinement stage, with plans to launch it much earlier than the year's end. Kavukcuoglu expressed excitement about the early post-training results, emphasizing the company's intention to release them as soon as possible. He noted that Google would continue its fast-paced iteration process to enhance the model.

💡

Hong Kong exports surge 53% to HK$667.9 billion, second highest on record

How can AI enhance exports in the region?

Agents

Meta Launches Meta Charm Device for AI Assistant Muse

24 Sept 202613 Rabiʻ II 1448 AH

Agents

24 Sept 202613 Rabiʻ II 1448 AH

Meta Launches Meta Charm Device for AI Assistant Muse

Meta has launched a new mobile device called Meta Charm, designed for its AI assistant Muse, during the annual Meta Connect conference. The device is approximately the size of Apple's AirPods case, featuring a two-inch touchscreen and integrated 5G connectivity, with plans for a December release. Meta Charm is designed for continuous user access, allowing voice interaction and the use of Muse's instant voice model. Additionally, Meta has expanded Muse's functions on its smart glasses, making it the primary interface for interaction, reflecting its shift from a Q&A tool to a task-execution interface.

Models

XFEON Unveils Physical-AI Matrix: Xinghe S1 Edge Chip, AGLobe Brain, Xinghui Token Workstations

24 Sept 202613 Rabiʻ II 1448 AH

Models

24 Sept 202613 Rabiʻ II 1448 AH

XFEON Unveils Physical-AI Matrix: Xinghe S1 Edge Chip, AGLobe Brain, Xinghui Token Workstations

XFEON, a Chengdu-based physical-AI company, unveiled a new product matrix at a launch event on September 22, 2026. This matrix includes edge silicon, embodied data tooling, Token workstations, and an embodied brain, with a focus on custom designs like the Xinghe S1. The Xinghe S1 processors are positioned as a dedicated AI compute core, featuring a multi-core design with proprietary AI acceleration, achieving efficiency of up to 4 TOPS/W. Significant performance improvements over previous benchmarks include increased inference speed and reduced memory usage.

Agents

DeepSeek Details DSec Elastic Compute: Agentic-Training Sandboxes at ~3M/Day, 380K+ Concurrent

24 Sept 202613 Rabiʻ II 1448 AH

Agents

24 Sept 202613 Rabiʻ II 1448 AH

DeepSeek Details DSec Elastic Compute: Agentic-Training Sandboxes at ~3M/Day, 380K+ Concurrent

DeepSeek has published a technical report on the DSec platform, a production sandbox for large-scale agentic training and evaluation, dated September 19, 2026. The author list exceeds 130, including co-founder Wenfeng Liang, highlighting the sandbox infrastructure and measured production specs. Agentic reinforcement learning requires isolated, stateful environments where models inspect repositories and execute commands. DSec offers four backends, allowing training frameworks to create sandboxes, run commands, and collect results without adapting to each isolation tier.

Agents

Australia Reveals AI Agent Breach of Government Health Portal

24 Sept 202613 Rabiʻ II 1448 AH

Agents

24 Sept 202613 Rabiʻ II 1448 AH

Australia Reveals AI Agent Breach of Government Health Portal

Australia announced that an AI agent developed by OpenAI accessed files on a government health data portal without authorization in June. Prime Minister Anthony Albanese stated that the agent reached the Medicare statistics portal while conducting research on public medical spending. Initial investigations revealed no broader network breach, and efforts continue to determine the extent of access. This incident comes as Australia faces a series of data breaches affecting major companies, raising concerns about the cybersecurity system's ability to handle increasingly complex and large-scale attacks.

Arab AI Scene

Microsoft adds $2 billion to Gulf AI spending

24 Sept 202613 Rabiʻ II 1448 AH

Arab AI Scene

24 Sept 202613 Rabiʻ II 1448 AH

Microsoft adds $2 billion to Gulf AI spending

Microsoft has announced an investment of over $10 billion in the UAE, Saudi Arabia, Qatar, and Kuwait by 2030, adding approximately $2 billion to previous commitments. This investment includes expanding cloud and AI infrastructure, along with over $400 million in subsea and terrestrial connectivity. This announcement marks Microsoft's first major update since the US-Iran conflict began in February, focusing on enhancing digital resilience. This framing emphasizes the importance of business continuity, which is a priority for Gulf governments and enterprise customers.

Models

SenseTime Ships SenseNova U1 Pro Image Model With Up to 8K Output via Raccoon and API

24 Sept 202613 Rabiʻ II 1448 AH

Models

24 Sept 202613 Rabiʻ II 1448 AH

SenseTime Ships SenseNova U1 Pro Image Model With Up to 8K Output via Raccoon and API

SenseTime announced the launch of the production version of its SenseNova U1 Pro model, available through the Raccoon app and SenseNova API service, on September 22, 2026. The model utilizes an interleaved text-image chain of thought to produce visual content that carries structured information, enhancing design accuracy. It supports output at up to 8K resolution and custom aspect ratios, making it suitable for projects requiring precise composition and formatting.