Large Language Models (LLMs) – Overview and Latest News

A federal judge has certified a class action lawsuit against AI firm Anthropic, escalating a high-stakes copyright battle amid conflicting judicial rulings on fair use.

Google rolls out an AI agent to call businesses and enhances its Search AI Mode for subscribers with the powerful Gemini 2.5 Pro model and Deep Search.

xAI's Grok 4 tops benchmarks but fails in real-world tests, revealing evidence of overfitting, ideological bias, and critical security flaws.

Researchers already jailbroke xAI's new Grok-4 model just two days after launch, using combined 'Echo Chamber' and 'Crescendo' attacks to bypass safety guardrails.

Chinese AI startup Moonshot AI has released Kimi K2, a 1-trillion-parameter open-source model aimed at claiming market leadership for agentic AI.

AI researchers have unveil the Energy-Based Transformer (EBT), a new AI architecture for 'System 2' reasoning that promises better scaling but at a higher compute cost.

Former Intel CEO Pat Gelsinger launches the Flourishing AI (FAI) benchmark via Gloo, a new test grading AI on human values like character, meaning, and faith.

Missouri's AG has launched a 'consumer fraud' probe into Google, Meta, and OpenAI after their AI chatbots ranked Donald Trump unfavorably, sparking a free speech outcry.

The European Union has published its final voluntary AI Code of Practice, providing key guidelines on safety, copyright, and transparency ahead of the AI Act.

Anthropic proposes a 'Secure Development Framework' for frontier AI, urging public safety disclosures to tackle the industry's growing trust and safety crisis.

Apple's AI division suffers a major blow as top engineering manager Ruoming Pang defects to Meta, the latest move in an aggressive, high-stakes talent war.

Alibaba has released WebSailor, a new open-source AI web agent designed for complex reasoning, claiming state-of-the-art performance that closes the gap on proprietary models.

Huawei's AI lab denies claims from research group HonestAGI that its Pangu model is a copy of Alibaba's Qwen, based on a novel 'fingerprint' analysis.

Microsoft has open-sourced its popular GitHub Copilot Chat extension for VS Code, a key step towards a transparent, community-driven AI editor.

German firm TNG has released DeepSeek-TNG R1T2 Chimera, an open-source variant twice as fast as its parent model thanks to a new 'Assembly-of-Experts' merge technique.

A new study from resarchers of Amazon, Stanford, MIT, and others reveals major flaws in AI agent benchmarks, finding they can misestimate performance by up to 100%.

Mistral AI is offering governments sovereign, customizable AI solutions as a direct challenge to dominant, non-European tech giants.

A coalition of 45+ top EU firms like Airbus & Mistral AI are demanding a 2-year delay on the EU AI Act, citing unclear rules and risks to innovation.

Baidu has overhauled its search engine with Ernie AI, and launched MuseSteamer, a new AI video generator, escalating the global AI race and price war.

GitHub has empowered its Docs site with a new Copilot search, capping a period of major AI upgrades including Agent Mode and new premium models.

Apple is reportedly considering a major strategic shift, exploring AI models from OpenAI and Anthropic to power Siri after years of internal development setbacks.

A leaked internal memo from OpenAI's Mark Chen reveals a "visceral" reaction to Meta's aggressive talent raids. The AI leader is now scrambling to "recalibrate comp" to prevent more top researchers from defecting to Mark Zuckerberg's firm.

Baidu is making its powerful Ernie generative AI model open source, a strategic pivot that declares a 'war on pricing' and directly challenges proprietary rivals like OpenAI and DeepSeek in the escalating global tech race.

Meta is escalating its AI talent war with OpenAI, hiring eight researchers in one week as a direct response to internal setbacks with its Llama models.

An unreleased OpenAI research paper defining the stages of superintelligence has ignited a high-stakes battle with Microsoft, threatening to unravel the tech industry's most influential partnership over control of future technology and the very definition of AGI.

Google has released Gemma 3n, a new generation of open, mobile-first AI models that bring powerful multimodal capabilities to consumer devices with as little as 2GB of memory, empowering developers to build sophisticated, offline-capable applications.

Meta has further escalated its AI talent war, hiring influential reasoning expert Trapit Bansal from OpenAI and entering talks to acquire voice AI startup PlayAI. This new, more surgical strategy reveals a company under intense pressure to fix internal setbacks and close the technology gap with its rivals.

The highly anticipated launch of Chinese AI startup DeepSeek’s next-generation R2 model has been indefinitely stalled, a major setback for one of the country's...

Meta's apparent win in a major AI copyright case is a legal illusion. A federal judge's ruling, while favoring Meta, provides a clear roadmap for future lawsuits to succeed by focusing on market dilution, putting the entire AI industry on notice.

Anthropic has launched a new feature allowing users to build, host, and share AI-powered apps directly within its Claude chatbot, a strategic move to create a developer-friendly platform that challenges competitors with a unique cost-shifting billing model.