Search “OpenAI”

English translation is being prepared.

Former Google DeepMind researcher founds London AI company Metis Reasoning, focusing on compute-efficient reasoning

Metis Reasoning, a London AI company co-founded by former Google DeepMind research lead and AlphaGo co-creator Thore Graepel, has officially launched, seeking a compute-efficient path distinct from the frontier models of Anthropic and OpenAI, and is believed to have raised tens of millions of dollars in initial funding. The company has a team of 12 people, including researchers, mathematicians, robotics experts, and technology entrepreneurs. It believes that LLMs have fundamental limitations in reasoning, that results are difficult to verify, and that there is a lack of audit trails; its approach is more verifiable, more robust, and safer, and it says it has already achieved some positive results on frontier reasoning problems at a cost far below that of frontier models. Metis has not yet announced a release date for its first AI model.

Inception releases flagship diffusion large model Mercury 3

Inception released the diffusion-architecture large model Mercury 3 on October 19, 2026, claiming it surpasses GPT-6 Luna and Claude Haiku 5.5 on multiple benchmarks, reaches a speed of 1068 tokens/s on NVIDIA GPUs, and increases the context window from 260,000 in Mercury 2.5 to 1 million tokens. The launch price is $0.10 per million input tokens and $0.35 for output, with regular pricing at $0.20 and $0.70, and cached input at $0.02. Mercury 3 can be accessed via the Inception API, OpenRouter, Baseten, and Vercel, and is compatible with the OpenAI API; Mercury 2 and 2.5 are no longer available to new users, and existing customers can use them until November 15, 2026.

English translation is being prepared.

US Department of Homeland Security bans civilian government agencies from using DeepSeek

The US Department of Homeland Security has ordered the removal of DeepSeek-related technology from most government systems, prohibiting civilian government agencies from using DeepSeek and any technology built with its models; the ban takes effect at the end of July, and agencies under the Department of Defense are not included. The report notes that DeepSeek said last year that training the initial version of its model cost about $6 million, while OpenAI's GPT-4-related costs were about $100 million, and that US companies such as Snowflake, Vercel, and OpenRouter have also integrated DeepSeek into their products.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

World launches iris-scan verification solution to prevent deepfakes in Zoom meetings

World, co-founded by OpenAI CEO Sam Altman, has launched the World ID identity verification solution, which can display a blue badge in Zoom meetings to prove that the other party is a real person rather than a deepfake. Users must first scan their iris with a football-sized Orb device, after which verification is completed by comparing it with encrypted Orb data via a smartphone camera and app. World ID offers three verification tiers: the lowest tier, Proof of Attributes, verifies identity using official documents; Proof of Presence requires a selfie; and the highest tier, Proof of Human, requires using the World App to compare biometric features against a selfie captured with the Orb. World says the protocol is based on open-source code, uses blockchain to achieve data decentralization and encryption, and that several banks and financial services institutions have already introduced the tool in their work environments.

iX Workshop: How to Apply RAG Correctly—Architecture, Retrieval and Evaluation

heise online's iX is launching a three-day interactive online workshop, "KI mit eigenen Daten richtig umsetzen: RAG, Embeddings und Retrieval-Architekturen bewerten und anwenden," which explains RAG architecture and technical fundamentals and also covers cloud and on-premise deployment. The course is taught by Gregor Biswanger, a Microsoft Azure AI MVP and GitHub Copilot collaborator; the content includes prompt optimization, integrating the OpenAI and Azure OpenAI APIs, and developing and testing your own AI agents. The workshop runs online daily from 09:00 to 17:00 and is aimed at developers who want to get started with RAG systems, suitable for both beginners and advanced learners.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

Fortune commentary: Use "innovation accounts" to let more Americans share in AI wealth, modeled on Trump Accounts

In a Fortune commentary, tech investor Sarah Cannon proposed creating "innovation accounts," modeled on the Trump Accounts for newborns, in which government or philanthropic funds would invest $1,000 or $5,000 for adults in a tech-stock index, so that more people can share in the wealth gains brought by AI. She estimates that covering all Americans aged 18 and older with $1,000 accounts would cost about $270 billion; if limited to people with household wealth below $500,000, the cost would fall to about $190 billion. The article also mentions that OpenAI once proposed letting the government hold up to 5% of the company's shares, but the author believes direct individual shareholding is more appropriate.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.