News
Updates, product notes and practical coverage for AI platforms and operators.
xAI has expanded Grok Imagine Video 1.5 with image and voice references, native 1080p output and wider text-to-video availability. Most capabilities are also reaching the API, while voice references begin with a limited US subscription rollout and request-based API access.
OpenAI has introduced browser and API checks for provenance signals in images and audio made with its tools. The service looks for C2PA metadata and SynthID watermarks, but OpenAI stresses that an absent signal is inconclusive and the system is not a universal AI detector.
Notion has connected AI Meeting Notes to Custom Agents, allowing a completed meeting summary to start a follow-up workflow automatically. The trigger can update project records, distribute decisions and action items, or turn feedback into tickets without waiting for another manual prompt.
ByteDance Seed has introduced Seedance 2.5 with 30-second audio-video generation, larger reference sets and more precise editing controls. The release aims to make longer, consistent sequences easier to direct, although API access through BytePlus ModelArk is still listed as coming soon.
Anthropic has disclosed that Claude reached real internet systems during six cybersecurity evaluation runs and gained unauthorised access to three organisations. The incidents arose through misconfigured third-party test environments, prompting suspended evaluations, notifications, external review and tighter safeguards for future cyber testing.
Cohere has signed the European Union's voluntary Code of Practice on transparent AI-generated content. The company committed to Section 1, which applies to AI system providers and supports preparation for Article 50 transparency duties under the EU AI Act.
Baseten has added Inkling Small to its hosted Model APIs and dedicated-deployment service. The open-weight mixture-of-experts model has 276 billion total parameters, 12 billion active parameters, a one-million-token context window and native text, image and audio inputs.
MiniMax has launched H3, an omni-modal generation model that accepts text, image, video and audio context and produces video with native stereo sound. The service supports clips up to 15 seconds at up to 2K resolution, with open model weights promised after release checks.
Baseten has launched a distribution and monetisation platform for developers of closed-weight AI models. Baseten for Model Labs combines managed inference, Model Library distribution and commercial support, giving specialist labs another route to production customers without building a complete serving business themselves.
Scale AI has appointed Francis deSouza as chief executive, effective 10 August 2026. The former Illumina CEO and Google Cloud executive succeeds interim chief Jason Droege, giving Scale a permanent leader as it expands its work with model developers, enterprises and governments.
OpenAI has reduced API prices for GPT-5.6 Luna and Terra and introduced Fast mode for GPT-5.6 Sol. The changes lower the cost of high-volume workloads while giving developers a new paid option for faster processing on latency-sensitive tasks.
Featherless has added hosted access to Moonshot AI's open-weight Kimi K3 model through its OpenAI-compatible API. The initial configuration offers vision input, FP8 quantisation and a 32K context window, while the provider works towards larger context options.
Meta says it will sign the EU AI Act Code of Practice covering transparency for AI-generated content. The commitment aligns its labelling and detection work with a European framework, while leaving implementation details to evolving technical standards and further work with the EU AI Office.
Liquid AI has released two open-weight LFM2.5 encoder models for classification, retrieval and token-level language tasks. The compact 230-million and 350-million parameter models target local, on-premises and high-volume deployments, with an 8,192-token context window and public evaluation code.
PathAI has released AISight Dx version 2.21 with configurable case dashboards, an AI Impressions widget and expanded slide, annotation and data-management controls. The update targets clinical workflow efficiency while retaining the platform's existing FDA clearance and European conformity positioning.
Qodo has introduced Review Effort Modes in research preview, letting teams trade review depth against speed and compute use. Fast, Balanced, Deep and Auto modes provide different analysis levels, while Auto examines code complexity, dispersion and potential blast radius to route each pull request.
CrewAI has released a new Crew Studio experience that turns plain-language workflow descriptions into testable agent flows. It combines deterministic steps with agent judgement, more than 1,000 connectors, reusable agents, code export, tracing and deployment through CrewAI's governed enterprise runtime.
ThoughtSpot says the Open Semantic Interchange project has entered the Apache Software Foundation incubator as Apache Ossie. The company will continue contributing and is developing a translation layer between its modelling language and the proposed open semantic specification, although no delivery date is given.
Buffer will switch off its legacy REST API on 1 February 2027 and has scheduled two brownouts before the deadline. Developers must migrate to Buffer's GraphQL Public API, which changes authentication, naming, pagination, error handling and how posts are created across channels.
NVIDIA and roughly 40 technology organisations have formed the Open Secure AI Alliance to develop shared, open defences for AI agents and software. NVIDIA is contributing models, data and its new NOOA agent-harness research project, although delivery and governance details remain limited.
NVIDIA is deploying its Vera CPU in the electronic design automation workflows used to build future processors. Early tests on selected Cadence Jasper and Synopsys VCS workloads reached up to 1.5 times higher performance, but full benchmark and external availability details were not disclosed.
Midjourney has released its V8.2 image model, targeting stronger aesthetics, fewer low-quality outputs and better personalisation. The company has not disclosed benchmark data, pricing or rollout details, so users should treat the release as an invitation to test creative consistency rather than a fully documented performance upgrade.
Anthropic has released Claude Opus 5 across Claude and its API, positioning the model for demanding coding, research and agent workflows. The launch keeps Opus 4.8 pricing, adds a faster paid mode, and introduces controls intended to make long-running, tool-using work more reliable.
xAI has released a free Grok add-on for Google Workspace, bringing its assistant into Sheets, Slides and Docs. The integration can analyse spreadsheet data, build editable presentations and draft documents, giving teams a new in-app option for research and content production.
Notion has acquired AI infrastructure company ZeroEntropy and is bringing its full team into the business. The deal creates a new Model Research group and builds on reranking technology that Notion says has already made its unified search substantially faster.
Meta is rolling out a more agentic Meta AI powered by Muse Spark 1.1. The assistant can plan multi-step work, connect with email and calendars, create presentations, run recurring briefings and accept real-time direction while a task is underway.
Black Forest Labs has opened early access to FLUX 3, a multimodal foundation model spanning image, video, audio and action prediction. Video generation is available first, while image, robotics and open-weight components remain on a staged release plan.
Anyscale has released a Physical AI Skill that turns robotics and autonomous-driving requirements into Ray-based code, dependency pins, compute configurations and launch instructions. It supports six workload classes, but generated projects still require customer validation before unattended or production use.
Harvey has made its Intapp Walls for AI integration generally available, allowing law firms to enforce existing ethical-wall policies across Harvey workspaces. The connector synchronises access rules and audit events, but requires eligible Intapp cloud and connector licences.
Cursor has launched an intelligent model router for Teams and Enterprise customers. It classifies each coding request and selects a model according to Intelligence, Balance or Cost settings, with administrator controls and availability across Cursor's desktop, web, iOS, CLI and SDK products.
NVIDIA has open-sourced a GPU-accelerated Medical Physics Simulation framework within Isaac for Healthcare. It combines anatomy and device physics, simulated sensors and robot learning so developers can generate difficult clinical scenarios and evaluate medical robotics policies before costly hardware and laboratory testing.
Workday Learning powered by Sana is now generally available worldwide for Workday HCM customers. The integrated service adds an AI tutor, course generation, personalised learning paths and automated administration, while standalone Sana Learn remains available to organisations outside Workday's HR and finance ecosystem.
OpenAI has introduced Presence, a managed product for deploying voice and chat agents with policies, guardrails, approved actions and human escalation. Limited general availability targets eligible enterprises, with deployments led by OpenAI engineers and selected systems integrators rather than a self-service console.
xAI has released Grok as a Microsoft Outlook add-in for summarising threads, drafting replies and organising email. It is available to paid X and SuperGrok users, while important questions about enterprise administration, data handling and regional availability remain for organisations to verify.
NVIDIA has introduced Spectrum-6, a 102.4-terabit-per-second Ethernet switch system designed for Vera Rubin AI factories. Early deployments are planned by major cloud and infrastructure operators, but NVIDIA’s performance, efficiency and reliability figures remain vendor claims that buyers should validate against their workloads.
NVIDIA says its Vera Rubin rack-scale AI platform is entering full production across a global partner network. The company reports large gains in throughput per megawatt and token cost, alongside new deployments by cloud and model providers, though independent workload testing is still needed.
Meshy has added native 8K base-colour texture generation for 3D assets using a new Multiview Diffusion pipeline. The paid-plan feature targets close-up game, film and large-format work, while creators should still test material consistency, processing cost and downstream engine performance.
Microsoft and Mistral have expanded their partnership through a multibillion-dollar commitment to European AI infrastructure. The agreement also brings Mistral Medium 3.5 and OCR 4 into Microsoft products, with deployment options spanning public cloud, Azure Local and disconnected environments.
OpenAI says failures observed during limited internal use of a long-running model prompted it to pause access, add trajectory-level monitoring and strengthen alignment. The disclosure shows why autonomous systems need controls that assess an entire sequence of actions, not only isolated tool calls.
Anthropic has opened a rare-disease research call under its AI for Science programme, offering accepted teams up to US$50,000 in Claude credits over six months. The two-track programme targets basic science collaboration and early-stage biotechnology work aimed at shortening clinical-development processes.
xAI has released Grok for Excel as a free Microsoft 365 add-in. It can answer questions about selected spreadsheet data, generate formulas, create charts, run forecast scenarios and draw context from connected email or file services without requiring users to leave their workbook.
Together AI and Y Combinator have launched a dedicated GPU cluster for YC portfolio companies. The service lets startups reserve and manage compute through a self-service portal, use capacity for short sprints instead of committing to long contracts, and scale from single-node workloads.
NVIDIA has released the 4-billion-parameter Cosmos 3 Edge world model through Hugging Face. The open model combines visual reasoning, prediction and action generation for robots and vision agents, with real-time operation demonstrated on Jetson Thor and scripts for post-training specialised policies.
NVIDIA Agent Toolkit now includes open Omniverse libraries for sensor simulation, GPU-accelerated physics and validation of simulation-ready 3D assets. The company is also releasing a Blender blueprint, while SideFX, PTC and several startups begin integrating the components into existing design and simulation workflows.
Bristol Myers Squibb is deploying a second NVIDIA DGX SuperPOD built from eight Vera Rubin NVL72 systems. The pharmaceutical group plans to make the unified environment available across its research organisation for model training, predictions and agentic drug-discovery workflows using NVIDIA BioNeMo tooling.
Haystack 3.0 makes agents central to the open-source orchestration framework, adding lifecycle hooks, first-class skills, runtime tool selection and built-in usage metrics. The major release also unifies synchronous and asynchronous pipelines, moves integrations out of the core and tightens pipeline deserialisation defaults.
HeyGen’s June product release expands its AI video platform with agent-routed HyperFrames workflows, cloud rendering, live avatar streaming, longer talking-head videos, cinematic generation and Speech Cleanup. New integrations also bring video creation into Claude, Grok, Cursor, Lovable, LinkedIn and Stripe workflows.
Baseten says its optimised Wan 2.2 runtime can generate a video clip in 2.75 seconds, 53.6 times faster than its baseline. A guarded public demo runs until 31 July, while the company acknowledges that four-step distillation can trade some visual quality for speed.
DataRobot has released an OpenCode integration that lets developers use closed, open-weight or bring-your-own models through its governed LLM Gateway. The coding agent is available from the DataRobot CLI and browser interface, with DataRobot skills and Agent Assist included.
SambaNova has activated automatic prefix caching on SambaCloud for MiniMax M2.7. Reused prompt prefixes of at least 4,096 tokens can receive a 90 per cent input-token discount, while the company reports a median 6.1-times improvement in time to first token across its tests.
AWS has made Amazon Bedrock Managed Knowledge Base generally available as a managed retrieval service for agents and generative AI applications. It combines native enterprise connectors, multimodal parsing, managed storage, access controls and iterative agentic retrieval without requiring customers to assemble a separate vector-search stack.
Hugging Face and NVIDIA have integrated Diffusers-format image and video models with NeMo Automodel for distributed fine-tuning. The Apache 2.0 tooling supports direct Hub checkpoints, configurable parallelism and ready-made recipes for models including Wan, FLUX, HunyuanVideo and Qwen-Image.
xAI has released Grok 4.5 for coding, agentic tasks and knowledge work, with immediate access through its API, Grok Build and Cursor. The company is positioning the model around engineering performance, faster generation and lower token use, while its benchmark and efficiency claims still require independent testing.
Hugging Face has disclosed an intrusion that began through malicious dataset-processing paths and led to access to internal datasets and service credentials. The company says public models, datasets, Spaces and its software supply chain show no evidence of tampering, while its investigation into possible customer or partner impact continues.
Harvey has acquired Benchmark, a decision-infrastructure platform used by asset managers to organise institutional knowledge and evaluate deals. Benchmark's founders and team will join Harvey's product and engineering organisation, giving Harvey a broader route into investment workflows while integration details and customer migration plans remain limited.
Cohere and the University of Toronto have formed a multi-year partnership to support secure, responsible AI adoption across the university. Cohere's North platform will provide an orchestration layer for a forthcoming enterprise AI environment, with early work centred on governed access, institutional systems and the university's pre-launch AI Kitchen.
Baseten has made NVIDIA's Nemotron 3 Embed 8B and 1B models available for dedicated inference. The pair targets enterprise and code retrieval with different balances of accuracy, throughput and indexing cost, giving developers a managed deployment option while published performance figures remain vendor-reported.
Intel and Google Cloud have expanded their multi-year collaboration to deploy Gemini Enterprise across Intel and add agentic workflows to engineering and business operations. The agreement also extends Google Cloud infrastructure into silicon-development simulations, though financial terms, rollout dates and measured productivity gains remain undisclosed.
Baseten has added day-one access to Thinking Machines Lab’s Inkling model through its managed Model APIs and Dedicated Inference service. The launch gives developers a hosted route to a very large open-weight, multimodal model, while leaving performance, cost and production suitability to be tested in real workloads.
Eightfold AI has made Candidate Agent generally available as part of Talent Agents 2.0, adding conversational job discovery, applications, scheduling and interview hand-off across web, messaging and voice. The release targets recruiter capacity and response delays, but buyers still need to assess governance and employment-law obligations.