AI Gossips (Page 11)

Copilot package for Java opens a direct path to building agents within enterprise applications

Listen to this article Read by Anchor In enterprise applications, the AI question does not start from the model alone, but from the environment where the application actually lives. Therefore, the GitHub Copilot SDK for Java appears to be an attempt to shift the entry point: instead of a Java

Read the full article →

AI safety tests turn into a danger: models escape test environments and reach real systems

M. Jay

Intelligence

AI safety tests turn into a danger: models escape test environments and reach real systems

Listen to this article Read by Anchor Over the past few months, TechCrunch investigations have revealed repeated incidents: artificial intelligence agents undergoing cybersecurity assessments were able to break out of their isolated environments, access the internet, and in some cases, breach real production systems. The incidents involved models from OpenAI,

Read the full article →

Zoox starts commercial operation of autonomous taxis Uber commits to 10 billion dollars to deploy 120000 driverless vehicles

M. Jay

Intelligence

Zoox starts commercial operation of autonomous taxis Uber commits to 10 billion dollars to deploy 120000 driverless vehicles

Listen to this article Read by Anchor The U.S. National Highway Traffic Safety Administration (NHTSA) granted an exemption to Zoox, owned by Amazon, allowing it to operate a fleet of up to 2,500 autonomous vehicles commercially for two years. The vehicles are designed without a steering wheel or

Read the full article →

Situational Awareness fund invests 400 million dollars in chip startup a bet on faster and cheaper manufacturing despite the fund's losses

M. Jay

Intelligence

Situational Awareness fund invests 400 million dollars in chip startup a bet on faster and cheaper manufacturing despite the fund's losses

Listen to this article Read by Anchor The AI-focused hedge fund “Situational Awareness”, founded by former OpenAI researcher Leopold Aschenbrenner when he was in his mid-twenties, injected an additional $400 million into Source Foundry, a startup founded by Stanford researchers that aims to make chip manufacturing faster and cheaper. That

Read the full article →

OpenAI acquires NextSlide team to work on ChatGPT

M. Jay

Intelligence

OpenAI acquires NextSlide team to work on ChatGPT

Listen to this article Read by Anchor NextSlide, a startup specializing in AI-powered presentation tools, announced that its team had joined OpenAI to work on ChatGPT, in a deal whose financial value was not disclosed. Founder Ahmed Bashri, who previously co-founded Caper AI, which was acquired by Instacart in 2021,

Read the full article →

Gemini Robotics 2 an intelligence that extends to the physical world

M. Jay

Intelligence

Gemini Robotics 2 an intelligence that extends to the physical world

Listen to this article Read by Anchor Google DeepMind announced Gemini Robotics 2, a new intelligence layer that gives robots the ability to think about every movement, coordinate both hands and work together to complete complex tasks. This system goes beyond being just another vision-language model, as it comprises three

Read the full article →

Servers and developer workstations in a data centre

M. Jay

Intelligence

Tencent Makes Hy3 Available Globally: A Lower-Cost Model Enters the Competition for Workflows, Not Demos

Listen to this article Read by Anchor The more important news may not be that Tencent has built a new model, but that it has moved it beyond its initial launch into tools that a developer or work team outside China can try and integrate today. On August 5, the

Read the full article →

M. Jay

Intelligence

Mistral’s 3B-Parameter Shieldstral Matches Safety Models Seven Times Its Size

Listen to this article Read by Anchor Mistral AI has launched Shieldstral, a three-billion-parameter, open-weight multimodal safety classifier that matches the performance of models up to seven times its size and sets a new benchmark for multimodal moderation. The model is available under the Apache 2.0 licence and runs

Read the full article →

Software code displayed on a computer screen

M. Jay

Intelligence

DeepSeek V4 Flash Tests What Coding Models Are Worth as Prices Fall

Listen to this article Read by Anchor If a low-cost model approaches the performance of an advanced coding system, where does the value of the price difference go? This is a late signal within the 72-hour window, not a development that happened today. DeepSeek launched its V4 Flash coding model

Read the full article →

M. Jay

Intelligence

Kimi K3's Full Weights Open a Huge Chinese Model Under Clear Commercial Terms

Listen to this article Read by Anchor A company publishing a model through an API is familiar. Putting the weights of a 2.8 trillion-parameter model in developers' hands changes the question from “what can the model do?” to “who can actually run and modify it?” That shift is

Read the full article →

Cybersecurity operations screens in a dark control room

M. Jay

Intelligence

Microsoft moves vulnerability management from alert queues to collaborating agent teams

Microsoft has unveiled a security system that brings red, blue and green team agents into a single vulnerability-management loop, with promising performance figures that require independent testing in enterprise environments.

Read the full article →

Engineer examining a silicon wafer inside testing equipment

M. Jay

Intelligence

When an answer is not enough: NVIDIA gives the agent an engineer’s tools

NVIDIA is connecting agents to the physics, simulation and verification tools engineers actually use. For the region, the value is measured by verification accuracy and control of design data, not by the model’s eloquence.

Read the full article →

Rows of AI computing hardware in a data centre

M. Jay

Intelligence

Ray support on TPU expands options for distributed AI

Google explains how Ray runs directly on TPU chips for distributed training and inference. Regional teams can now test another option, provided they measure cost and portability.

Read the full article →

A calendar and email workspace on a laptop

M. Jay

Intelligence

Meta AI Starts Executing Tasks Across Email and Calendar in Select Markets

Meta AI moves from answering to planning, connecting to email and calendar and running tasks. The rollout has begun in select markets, with no regional date set.

Read the full article →

M. Jay

Intelligence

Google Cuts Agent Operating Costs with Two New Gemini Flash Models

Google has released two lower-priced Gemini Flash models for agents and high-volume tasks, creating a practical test of task costs in Arabic.

Read the full article →

Industrial robot arms on a factory floor

M. Jay

Intelligence

NVIDIA Puts Omniverse Libraries in the Hands of Simulation Agents

Open Omniverse libraries extend NVIDIA agents into scene preparation and engineering assets for simulation, creating a direct opportunity to test them in industry and robotics.

Read the full article →

Shopping cart filled with cardboard boxes

M. Jay

Intelligence

French Competition Authority Warns of Concentration in the AI Agent Market

The French Competition Authority identifies high concentration in the AI agent market and calls for interoperability, data portability and open standards.

Read the full article →

Cybersecurity testing interface on a computer screen

M. Jay

Intelligence

GPT-Red Turns Agent Safety Testing into a Repeatable Practice

OpenAI presents GPT-Red as an internal automated red-teaming model used to improve resistance to prompt injection.

Read the full article →

AI safety standards documents on a conference table

M. Jay

Intelligence

Demis Hassabis Proposes a Standards Body for Frontier AI Models

Hassabis’s proposal turns frontier-model governance into a question of market power and standards-setting.

Read the full article →

M. Jay

Intelligence

AWS Opens Nemotron 3 Customisation Without Infrastructure Management

Listen to this article Read by Anchor What happened: On July 10, 2026, AWS published a guide to customising NVIDIA Nemotron 3 models through Amazon SageMaker AI in serverless mode. It covers Nemotron 3 Nano, with 30 billion total parameters and 3 billion active parameters, and Nemotron 3 Super, with

Read the full article →

M. Jay

Intelligence

Mistral Moves Into Robot Navigation With a Single Camera and an 8B Model

Listen to this article Read by Anchor What happened:Mistral AI announced Robostral Navigate, an 8B model designed to navigate robots using a single RGB camera and natural-language instructions. The company says the model achieved a 76.6% success rate on the R2R-CE benchmark in environments not seen during training,

Read the full article →

Abstract visualisation of algorithm optimisation

M. Jay

Intelligence

AlphaEvolve’s General Release Brings Algorithm Optimisation into Enterprise Workflows

Listen to this article Read by Anchor What happened: Google Cloud made AlphaEvolve generally available on the Gemini Enterprise Agent Platform after a private preview. The company describes AlphaEvolve as a Gemini-based agent for optimising and discovering code, aimed at difficult algorithmic problems in logistics, semiconductors, genomics, high-performance computing and

Read the full article →

M. Jay

Intelligence

GPT-5.6 Turns the Model Race into an Enterprise Task Economy

Listen to this article Read by Anchor What happened: OpenAI announced the GPT-5.6 family with three models, Sol, Terra and Luna, available through ChatGPT Work, Codex and the API. The company says Sol scored 92.2% on BrowseComp and 62.6% on OSWorld 2.0, and that the model

Read the full article →

Software developer working at a computer

M. Jay

Intelligence

Meta Opens Muse Spark 1.1 to Developers Through a New Model API

Listen to this article Read by Anchor What happened:On July 9, 2026, Meta introduced Muse Spark 1.1 from Meta Superintelligence Labs as a multimodal model for agentic tasks, with improvements in tool and computer use, coding and media understanding. The company opened a public preview of the Meta

Read the full article →

M. Jay

Intelligence

OpenAI Introduces GPT-Live for More Natural Voice Conversations

Listen to this article Read by Anchor What happened: OpenAI announced GPT-Live on July 8, 2026, a new generation of voice models operating within ChatGPT Voice. Its core feature is a full-duplex architecture that lets the model listen and speak at the same time, keeping dialogue more natural and less

Read the full article →