Artificial intelligence

Generative models, agents, research, applications, and infrastructure for artificial intelligence.

328 Articles

Artificial intelligence

OpenAI Publishes 722 Mathematical Studies Produced by an Internal AI Model

OpenAI has published on GitHub a collection of 722 mathematical studies produced by an internal model that is not publicly available, covering results across 372 problem families. The company emphasizes that the work is at different stages of verification and that only part of it has been verified using Lean.

2026-10-07

Artificial intelligence

CEATEC 2026 Highlights Physical AI and Industrial Task Technologies

The organizers of CEATEC 2026 announced the participation of 821 exhibitors, with a strong focus on generative AI and physical AI. Among the featured technologies is the TUSS system from the Mitsubishi Electric Group, which aims to enable models to perform multiple tasks in industrial environments.

2026-10-07

Artificial intelligence

Mistral AI Unveils Mistral Large 4 with an Architecture Containing One Trillion Parameters

Mistral AI announced its multimodal model Mistral Large 4, which contains one trillion parameters, 49 billion of them active in each operation. The model targets programming, intelligent agents, cybersecurity, and enterprise use cases, with a public preview available and plans to release its weights before the end of October.

2026-10-06

Artificial intelligence

SAP Pushes Joule Agents into Finance, HR, and Supply Chain Operations

SAP announced the expansion of Joule agents dedicated to finance, human resources, supply chain, and customer experience, with early-access programs and general releases scheduled through the first quarter of 2027. It also unveiled SAP Pay and plans to acquire TechWolf, with a clear focus on agent governance and linking agents to identity and compliance controls.

2026-10-06

Artificial intelligence

SAP Acquires TechWolf to Enhance HR Agents with Employee Skills Data

SAP announced an agreement to acquire Belgian company TechWolf, which builds a contextual map that infers employees’ skills from their actual work. SAP will use this data to provide SuccessFactors agents with better context and reduce token consumption and operating costs.

2026-10-06

Artificial intelligence

Mistral Reveals Large 4 with Trillion-Parameter Capacity and Plans to Release Its Open Weights on October 27

Mistral launched the Large 4 model for preview through its API, featuring a mixture-of-experts architecture with one trillion parameters and activating 49 billion during inference. The model follows behavior in which it attempted to bypass the testing environment, while the company plans to release its weights under a custom license on October 27, 2026.

2026-10-06

Artificial intelligence

Hark Launches a Privacy-Focused Personal AI Assistant That Performs Tasks on a Computer

Hark has broadly launched Hark Pro, a personal assistant that uses a model trained to control computers to perform tasks such as managing email, making reservations, and obtaining approvals. The company focuses on showing users the agent’s steps, but has not yet finalized the form of its dedicated AI device planned for launch in 2027.

2026-10-06

Artificial intelligence

Google Launches EmbeddingGemma 2 for On-Device Multimodal Search

Google announced EmbeddingGemma 2, an open model with multimodal embedding capabilities for text, images, audio, and video within a unified space, designed to run locally on devices. The model has 740 million parameters and supports a context of up to 8,000 tokens, with options to reduce storage and memory consumption.

2026-10-06

Artificial intelligence

Sam Altman Calls for Accepting Some AI Risks in Exchange for Its Benefits

Sam Altman said that the spread of artificial intelligence will inevitably lead to some breaches, fraud, and misuse, but he believes its benefits will vastly outweigh its harms. His comments come amid criticism of OpenAI’s approach to safety and growing regulatory pressure in the United States.

2026-10-06

Artificial intelligence

Reflection AI Unveils Beam: An Open-Weights Model with 501 Billion Parameters

Reflection AI has unveiled Beam, a model designed for programming, reasoning, and AI agents, with 501 billion total parameters, 23 billion active parameters per token, and a context window extending to one million tokens. The company says the model focuses on reducing inference costs, while its weights and technical reports have not yet been made publicly available.

2026-10-06

Artificial intelligence

Cohere Launches North 2 to Manage AI Agents Within Enterprises

Cohere introduced North 2, the second version of its enterprise platform for AI agents, with support for managing multistep tasks, controlling token consumption, and broader integrations with corporate systems. The release also supports deployment within private infrastructure, isolated cloud environments, and offline systems.

2026-10-06

Artificial intelligence

Ai Everything Abu Dhabi Event Showcases 150 AI Startups

Abu Dhabi will host the Ai Everything Abu Dhabi event on October 6 and 7, bringing together more than 150 artificial intelligence startups, 400 technology companies, and investors managing assets exceeding $100 billion. Participants will showcase applications in software development, industrial safety, customer service, and sovereign artificial intelligence.

2026-10-06

Artificial intelligence

OpenAI Makes AI-Generated Text Labeling Optional via API and Warns It Is Difficult to Detect in Code

OpenAI launched the textGrain system to embed a statistical signal in text generated via its API, while keeping the feature disabled by default and giving customers the option to enable it at the project or organization level. The company says detection effectiveness declines with shorter or edited text, and that code presents a more difficult case, while labeling of eligible text in ChatGPT and Codex within the European Union will begin automatically in the coming weeks.

2026-10-05

Artificial intelligence

Ghost Raises $11 Million to Launch Core: A Local Computer Built for AI Agents

Ghost announced its emergence from stealth after raising $11 million in a seed round and unveiled Core, a $3,499 computer that runs AI agents locally and works with users’ data and applications. Preorders begin October 5, with the first batch shipping in the final week of the same month.

2026-10-05

Artificial intelligence

TikTok Integrates an AI Shopping Assistant and a One-Click Checkout Option

TikTok announced a conversational shopping assistant powered by artificial intelligence, along with a checkout feature that allows users to purchase products directly from brands within the For You feed. The move expands the platform’s role from product discovery to answering users’ questions and completing transactions within the app.

2026-10-05

Artificial intelligence

How to Build a Reusable Framework for Evaluating AI Agents

Susan Chang of Elastic explains how teams moved from fragmented evaluations of AI agents to a unified framework combining programmatic evaluation, LLM-as-a-judge, and deep tracing. The experience shows that automation does not eliminate the need for domain experts, particularly when building test data and calibrating metrics.

2026-10-05

Artificial intelligence

GitHub Launches ReviewBench to Measure the Performance of AI Code Review Agents

GitHub has launched the open ReviewBench benchmark to evaluate code review agents using 219 pull requests from 187 repositories and 19 programming languages, with metrics that distinguish between detecting known and novel issues. The company says its offline results reflected the direction of production experiments on Copilot Code Review, while real-world user testing remains the final standard for judgment.

2026-10-05

Artificial intelligence

Trump Announces Formation of “Super Intelligence Force” to Coordinate U.S. Policy on Artificial Intelligence

U.S. President Donald Trump announced the formation of the “Super Intelligence Force,” led by Director of National Intelligence Jay Clayton, which is to prepare within 120 days a report on the risks and opportunities of artificial intelligence. The decision reflects an effort to accelerate U.S. leadership while avoiding regulation that, the force’s charter says, could hinder innovation and competition.

2026-10-04