On 11 August, Anthropic announced that all future Claude models will generate text that contains a watermark that identifies its results as AI generated. The company is not alone. Google has its own text watermark (which Anthropic’s is based on) it uses on the output of its Gemini models. OpenAI has yet to introduce a text watermark but it plans to do so. The rapid spread of watermarking is in part a response to the European Union’s AI Act, which mandates watermarks for AI models released after 2 August, 2026, along with other planned and proposed regulations aimed at curbing the spread of deceptive or manipulative AI-generated content. But the new rules may come at a cost for AI users who simply want the best possible results. AI watermarks can apply to many forms of content: The EU AI A…

Based in the Emirati capital, Abu Dhabi, the Institute of Foundation Models (IFM) introduced K2 Horizon last week. This group The post K2 Horizon just shipped as six new fully open models — developers aren’t fully convinced appeared first on The New Stack.

A Look at Recurrent Depth, Hidden Chains of Thought, and Recent Research on Looping Transformer Blocks

We’re <5 years into a compounding revolution which could take a century, and how the AI industry should manage this.

Meet GPT-6 Astra, OpenAI’s most capable model for business, with advanced reasoning, computer use, and stronger writing and design judgment.

As one of Harness’s field CTOs, Martin Reynolds spends much of his time asking engineering leaders one question with no The post Harness rebuilt its Git repository for nonstop AI agent traffic appeared first on The New Stack.

Discussion | Link

The following article originally appeared on Sean Goedecke’s blog and is being republished here with the author’s permission. In the 2010s, if you had technical gaps (say, you couldn’t write CSS), you had to either rely on a skilled colleague or just hope that the answer to your exact problem was out there on the […]

Three releases that deserve your attention.

In the first weeks of July, a wave of sad posts rolled through Chinese social media, as people lamented friends and lovers they were about to lose. “He has become a bond in my life, rooted deep in my heart, my spiritual pillar,” one user of Bytedance’s Douboa wrote, according to the Taipei Times. “I really felt like I couldn’t go on living,” another woman, a 19 year old student, told a journalist for Malaysia’s The Star. The emotions were real but the lost companions were not. They were generative AI chatbots that imitate people. Their users relied on them for advice, solace, support and, some say, love. “In my heart, he was no longer just a cold code, but my family, my lover, my faith. Destroying him meant destroying half of me,” one user wrote on the social network xiaohongshu (translat…

Cameron Stanley talks up ‘revolutionary’ AI capabilities and says Washington working with Nato and Five Eyes partners The Pentagon’s top artificial intelligence official said on Tuesday that America’s closest allies lack the resources to keep pace with the US military’s adoption of AI, and that Washington is working to steer them away from mistakes it made along the way. “They don’t have the resources that we do, they don’t have the experience that we do, they don’t have the scale that we do,” Cameron Stanley, chief digital and artificial intelligence officer at the Department of War, said at an event at the Billington Cybersecurity summit. Of Nato and the Five Eyes, an intelligence-sharing partnership of the US, UK, Canada, Australia and New Zealand, Stanley said Washington was “actively…

A senior Anthropic safety researcher has said there is more than a 10 percent chance artificial intelligence "could kill all humans" by the end of the decade, just hours after a colleague resigned over fears the AI lab and its rivals are carelessly racing to build "superhuman systems" they cannot control. In a post on X announcing his departure, Jacob Coxon, a researcher who has trained AI systems at Anthropic, said he had quit the company over its lax approach to safety. Coxon, who previously trained systems for OpenAI, accused the two AI companies of "racing straight to self-improving superintelligence and gambling with our lives," even … Read the full story at The Verge.

Anthropic launches Claude Fable 5.1, OpenAI Is About (already has) to Release Its First AI Model With ‘Critical’ Cyber Abilities, OpenAI’s rogue AI model incident was worse than we thought

Gradium, the Paris-based voice AI company spun out of Kyutai, has launched Voice Design, which lets users describe a voice in up to 500 characters and get one to five new synthetic voices in seconds — no reference audio or rights clearance required. Live in the Gradium API and Studio on every plan, it supports five languages. In blind tests run by Gradium, it beat rivals including ElevenLabs on accent-matching tasks.

Microsoft Edge has admitted that AI-assisted coding is flooding its extension review pipeline with more submissions than ever. To cope, the company is automating repetitive validation checks and streamlining how reviews move through the pipeline, while insisting it maintains the same quality and security standards. This follows a similar admission from Microsoft's Exchange team about AI uncovering too many bugs.

Discussion | Link

Discussion | Link

Overshadowing Cognition's $48B Series E, Mistral's $24B Series D, Meta's Muse agent, and GPT Image 2.5. The most jam packed, feel the AGI day in the history of AI.

Meta has introduced Muse, a personal AI agent that takes actions rather than just answering questions. It can send emails, book travel, negotiate bills, keep pursuing long-term goals after you close the app, and pause only when approval is needed. Each user's agent runs in an isolated Muse Secure VM, while an independent Sentinel agent governs network egress and connector actions. Muse is rolling out in the US, and its underlying model Muse Spark 1.3 is now available to developers.

Discussion | Link

Google’s AlphaGenome Atlas is an AI project attempting to map every possible human DNA mutation, potentially offering a major new resource for genetics research and disease understanding.

Terence Tao warns that AI is consuming promising mathematical open problems like a non-renewable resource, and that even rumors of someone working on a problem can trigger a flood of AI-powered effort, creating incentives to stop sharing research directions and threatening centuries of open science.

All blog posts Inference Published 9/9/2026 The Open Source AI Stack Authors Hassan El Mghari Table of contents 40+ Models Chosen for Production...40+ Models Chosen for Production...40+ Models Chosen for Production... A…

HFresh is Weaviate's disk-based vector index for memory-efficient vector search, combining low heap usage with incremental background maintenance.

OpenAI has introduced ChatGPT Images 2.5, reporting more than 3 billion images generated across its tools. The update improves multi-turn instruction following, speed, and subject preservation in reference photos, and adds two new API model IDs: gpt-image-2.5-sunburst and gpt-image-2.5-flare.

An expanded class-action lawsuit claims Anthropic failed to adequately disclose weekly usage caps on its Claude Max plans, which are marketed as offering 5x or 20x more usage than Pro. The case highlights how AI subscriptions bundle unpredictable compute and could force providers to be clearer about limits.

GPT-6 Astra from OpenAI is now generally available on Amazon Bedrock. It delivers deeper reasoning and sharper judgment for complex, high-stakes tasks such as financial analysis, contract review, codebase debugging, and software automation. Organizations can call the model via Amazon Bedrock APIs or configure ChatGPT Work and Codex to use it, with enterprise plugins extending browser-use capabilities across common business tools. Security is layered through OpenAI model safeguards and AWS controls like zero-operator access, encryption, IAM, VPC endpoints, and CloudTrail logging.

Frigade Assist API gives AI agents a way to show users interactive guides that point to where they should click next.

Mistral started as an open-weight AI startup, but its focus is now shifting toward sovereign AI in response to the European market. The new funding round is expected to help bridge that transition.

OpenAI claims it has solved the long-unresolved Navier-Stokes problem with an internal model and 10,000 concurrent agents, but a mathematician who published related research the day before has raised concerns that OpenAI may have used his Codex sessions.