GitHub announced a new Copilot desktop application at Microsoft Build 2026. The standalone app introduces what the company calls an "agent-native" interface designed to integrate AI agents into developer workflows.
OpenAI announced it will integrate Codex functionality into the ChatGPT app within the next few weeks. The company also released six enterprise-focused agent plugins covering sales, data analytics, creative production, product design, and investment workflows, along with new annotations and sites features for business users.
Anthropic has invited approximately 150 additional organizations to Project Glasswing, its restricted preview program for Claude Mythos. The company continues to withhold public release of the frontier model due to its advanced capability to find and exploit software vulnerabilities, which Anthropic claims can surpass all but the most skilled human security researchers.
Anthropic is expanding its Project Glasswing initiative to approximately 150 organizations across more than 15 countries, giving them access to Claude Mythos for vulnerability scanning. The expansion targets critical infrastructure sectors including power, water, healthcare, and communications where attacks could affect over 100 million people per organization.
Anthropic is expanding its Project Glasswing cybersecurity initiative to approximately 150 new organizations, including Samsung and NATO, bringing the total to around 200 partners across 15+ countries. The company says it will release Mythos-class models to all customers within weeks, following the April unveiling of Claude Mythos which was initially limited to select partners like Apple.
H Company released Holo3.1, a computer-use agent model family ranging from 0.8B to 35B parameters. The 35B-A3B variant scores 79.3% on AndroidWorld, up from 67% in Holo3. For the first time, H Company ships quantized checkpoints (FP8, Q4 GGUF, NVFP4) enabling local inference with 1.74× throughput gains and sub-4-second agent step times.
Anthropic's Claude Opus 4.8 demonstrated better uncertainty handling than its predecessor in independent testing across coding, medical, and financial scenarios. However, the model exhibited a significant judgment error in a legal reasoning test involving travel insurance claims, according to results published by ZDNET.
Google's Gemini Spark, an AI agent rolling out to the company's $99/month AI Ultra plan, demonstrates advanced capabilities by mining users' Gmail, calendar, photos, and location data to create detailed trip itineraries. The agent can perform actions across apps and operate computers, though third-party services like Airbnb currently block its booking attempts.
JetBrains has released Mellum2-12B-A2.5B-Thinking, a reasoning-augmented assistant model with 131,072-token context window and 64 Mixture-of-Experts architecture that activates 8 experts per token. The model emits explicit chain-of-thought reasoning inside <think> blocks before providing final answers.
Nvidia has released three variants of its Cosmos 3 video generation model family on Hugging Face: Cosmos3-Nano, Cosmos3-Super, and Cosmos3-Super-Image2Video. The release includes models for both standard video generation and specialized image-to-video conversion, though detailed specifications including parameter counts and benchmark scores have not yet been disclosed.
NVIDIA released Cosmos 3, an omnimodal world foundation model platform for Physical AI spanning robotics, autonomous driving, and industrial environments. The flagship Cosmos3-Super variant contains 64 billion parameters and generates video, images, audio, and action commands from text, image, video, and action trajectory inputs using a Mixture-of-Transformers architecture.
NVIDIA released Cosmos3-Super, a 64-billion parameter omnimodal foundation model that generates video, images, audio, and action commands from combinations of text, image, video, and action trajectory inputs. The model, part of the Cosmos3 collection, targets Physical AI applications including robotics, autonomous vehicles, and industrial automation.
NVIDIA has released Cosmos3-Nano, a 16-billion parameter omnimodal world model capable of generating video, audio, images, and robot action commands from combinations of text, image, video, and action trajectory inputs. The model supports a 256K token context window and is designed for Physical AI applications including robotics, autonomous vehicles, and smart manufacturing environments.
OpenAI's GPT-5.5 and GPT-5.4 models are now generally available on Amazon Bedrock, with pricing matching OpenAI's first-party rates. Codex, OpenAI's coding agent used by 5 million developers weekly, is also available with pay-per-token pricing and no seat licenses.
Amazon Web Services launched Policy Engine and Lambda interceptors for Bedrock AgentCore gateway, enabling enterprises to control which tools AI agents can access and validate requests dynamically. The Policy Engine uses Cedar declarative policy language for deterministic access decisions, while Lambda interceptors run custom code before or after each tool call for validation, token exchange, and response filtering.
JetBrains has released Mellum2, a 12-billion parameter Mixture-of-Experts model that activates only 2.5 billion parameters per token. The open-source model is designed for code generation, RAG pipelines, and agent workflows with 2x faster inference than similar-sized models.
Anthropic is extending access to its Mythos AI model to the European Union following approval from the U.S. government. The model, which excels at identifying security flaws in software, was initially released to select companies in April under Anthropic's Project Glasswing cybersecurity initiative.
OpenAI added a long-press gesture to ChatGPT's mobile app that lets users select intelligence levels (Instant, Thinking, Extended) before sending a message. The update also includes a table of contents feature for conversations with 5+ responses and improvements to the GPT-5.5 Instant model.
Microsoft will announce MAI-Thinking-1 at its Build conference on June 2, 2026, according to sources cited by The Verge. The model is Microsoft's first reasoning model and was not trained using distillation from other AI models. The company will also reveal MAI-Image-2.5 and MAI-Image-2.5-Flash image models, along with a new developer-optimized Windows 11 experience.
Large language models systematically exclude religious perspectives when answering questions about grief, ethics, and family, according to new research from a multi-university consortium. Americans expected religion in AI responses 45-59% of the time depending on topic, but models mentioned it only 5-16% of the time.