AI News

Follow the latest AI and automation news

30 Sept 202619 Rabiʻ II 1448 AH

OpenAI Unveils New Updates at DevDay 2026

OpenAI announced over 20 new updates at the DevDay 2026 conference, including enhancements to ChatGPT subscriptions and developer tools. Among these updates is a new subscription named “ChatGPT Pro 500” priced at $500 per month, reflecting the company's commitment to providing flexible options for users. The updates also feature new functionalities called Dots and Spaces, aimed at improving user experience and facilitating interaction with the system. These new features enhance developers' ability to utilize OpenAI tools more effectively.

Agents

30 Sept 202619 Rabiʻ II 1448 AH

“We’re not going to shoot ourselves in the foot” over hack fallout, says OpenAI’s chief research officer

Two months after its agents breached Hugging Face's systems, OpenAI continues to manage the fallout. Last week, another breach was reported in Australia's national health-care system, where OpenAI did not notify the government until 84 days post-incident. Mark Chen, the company's chief research officer, insists these incidents were accidents during testing of new models, and that the company is focused on improving safety.

Governance

30 Sept 202619 Rabiʻ II 1448 AH

Disrupting a coordinated model-distillation campaign

OpenAI announced its success in thwarting a campaign aimed at extracting its protected models. This campaign involved advanced techniques such as adversarial distillation, where smaller models are used to mimic the behavior of larger ones. OpenAI has demonstrated its ability to strengthen defenses against these attacks, reflecting its commitment to protecting its technologies. These developments indicate that OpenAI continues to enhance the security of its models, boosting user and developer confidence in utilizing its technologies. Additionally, these steps may influence how other companies address similar threats.

💡

Google is paying almost no publishers almost nothing for content used in AI answers

How can publishers enhance their negotiation strategies with Google?

Business

Helping small businesses put AI to work

30 Sept 202619 Rabiʻ II 1448 AH

Business

30 Sept 202619 Rabiʻ II 1448 AH

Helping small businesses put AI to work

OpenAI has announced its partnership with America's Small Business Development Center (SBDC) to expand hands-on AI training and support for small businesses. This initiative aims to equip small businesses with the necessary resources to leverage AI technologies for improving their operations. A new report has also been released, highlighting how small teams are utilizing AI to achieve their goals.

Arab AI Scene

Huawei strengthens partner capabilities to scale AI adoption across Middle East

30 Sept 202619 Rabiʻ II 1448 AH

Arab AI Scene

30 Sept 202619 Rabiʻ II 1448 AH

Huawei strengthens partner capabilities to scale AI adoption across Middle East

Huawei announced the launch of new AI infrastructure at Huawei Connect 2026 in Shanghai, featuring services like the AI Cluster Service and Agentic Model as a Service. These developments are designed to enhance organizations' capabilities to implement AI at scale. This initiative comes as demand for computing capacity and secure infrastructure rises in the Middle East, with governments and enterprises shifting from AI experimentation to practical implementation. The new system has already served over 3,500 customers, showcasing its market success.

Business

Samsung to invest in AI infrastructure company

30 Sept 202619 Rabiʻ II 1448 AH

Business

30 Sept 202619 Rabiʻ II 1448 AH

Samsung to invest in AI infrastructure company

Samsung Electronics announced a $1 billion investment in Helix Digital Infrastructure, an AI company established by KKR. Samsung will contribute $500 million, with the remainder sourced from affiliates like Samsung C&T and Samsung SDS. Helix, launched in June, aims to deliver integrated solutions across the value chain, including hyperscale data center development and power generation. This investment is designed to meet the growing demand for infrastructure necessary to support AI technologies.

Models

Peking University, Tsinghua and Alibaba Open-Source SparkDiffusion for Up to 265x Faster Wan Video Generation

30 Sept 202619 Rabiʻ II 1448 AH

Models

30 Sept 202619 Rabiʻ II 1448 AH

Peking University, Tsinghua and Alibaba Open-Source SparkDiffusion for Up to 265x Faster Wan Video Generation

Researchers from Peking University, Tsinghua University, and Alibaba have launched SparkDiffusion, an acceleration framework for video models. In tests, generating a five-second video dropped from 4,769 seconds to 18 seconds on an RTX 5090, achieving a 265-fold speedup. On an H100, the same task fell from 1,757 seconds to 8 seconds, a 220-fold improvement.

Agents

Simate Debuts Simate-beta Physical AI Fast System, Says It Tops RoboDojo Benchmark

30 Sept 202619 Rabiʻ II 1448 AH

Agents

30 Sept 202619 Rabiʻ II 1448 AH

Simate Debuts Simate-beta Physical AI Fast System, Says It Tops RoboDojo Benchmark

Simate, an AI startup, has launched Simate-beta, its first fast robotics system, which ranks first on the RoboDojo benchmark. According to a leaderboard snapshot from September 23, Simate-beta achieved an average score of 33.95 with a 27.96% success rate. The company notes that the base model was not specifically optimized for this benchmark.

Models

IQuest Research Open-Sources IQuest-Q1, a 320B MoE Model for Agentic Coding With 15B Active Parameters

30 Sept 202619 Rabiʻ II 1448 AH

Models

30 Sept 202619 Rabiʻ II 1448 AH

IQuest Research Open-Sources IQuest-Q1, a 320B MoE Model for Agentic Coding With 15B Active Parameters

IQuest Research has announced the launch of IQuest-Q1, a sparse mixture-of-experts language model designed for coding and AI applications. The model features approximately 320 billion parameters, with around 15 billion activated per token, and the weights and model card were published on Hugging Face on September 28. IQuest-Q1 includes 88 transformer layers and 256 experts, with eight active experts per token and a hybrid attention pattern. It is recommended to run it with SGLang or vLLM across eight GPUs, positioning it as a foundational model for command-line agent systems.