OpenAI restricts cybersecurity AI access following Anthropic's model controls

Meta released Muse Spark this week, its first major new AI model in over a year, marking a strategic shift from open-source Llama models to proprietary technology. The release comes nearly 10 months after Meta spent over $14 billion to hire Scale AI co-founder Alexandr Wang and establish Meta Superintelligence Labs. The critical question remains how Meta will monetize the model and compete with established rivals OpenAI, Anthropic, and Google.

Anthropic has declined to publicly release Claude Mythos Preview, its most capable AI model, citing critical cybersecurity risks. Instead, the company launched Project Glasswing, providing controlled access to 50+ organizations including AWS, Apple, Google, and Microsoft, along with $100 million in usage credits and $4 million in direct donations to open-source security initiatives.

Zhipu AI has released GLM-5.1, a freely available open-weight model designed for long-running programming tasks that achieves 58.4% on SWE-Bench Pro, edging out GPT-5.4 (57.7%) and Claude Opus 4.6 (57.3%). The model's core capability is iterative strategy refinement—it rethinks its approach across hundreds of iterations and thousands of tool calls, recognizing dead ends and shifting tactics without human intervention. However, GLM-5.1 trails on reasoning and knowledge benchmarks, scoring 31% on Humanity's Last Exam compared to Gemini 3.1 Pro's 45%.

YouTube is rolling out an AI avatar feature that lets creators generate digital versions of themselves for use in Shorts videos. The tool requires users to record a "live selfie" with face and voice data, generates clips up to 8 seconds long, and marks all AI-generated content with watermarks and digital labels.

Google has fully integrated NotebookLM, its AI research tool, into the Gemini app, allowing users to create notebooks with multiple source types directly within the chatbot. The feature supports PDFs, documents, website URLs, YouTube videos, and text input, with Gemini generating summaries, infographics, and audio overviews from uploaded sources.

Anthropic has launched Claude Managed Agents as a public beta, providing developers with a hosted platform to build and run autonomous AI agents without managing their own infrastructure. The service costs $0.08 per session hour on top of standard token pricing and is currently available exclusively on Anthropic's infrastructure.

Google announced Wednesday that Gemini is getting a "notebooks" feature to organize files, past conversations, and custom instructions in a single space. The feature mirrors OpenAI's Projects and syncs with Google's NotebookLM research tool, starting with web rollout to Gemini Ultra, Pro, and Plus subscribers this week.

Meta announced Muse Spark today, its first model release since Llama 4 a year ago. The hosted model is available in private API preview and on meta.ai with Instant and Thinking modes, benchmarking competitively against Anthropic's Opus 4.6 and Google's Gemini 3.1 Pro, though behind on Terminal-Bench 2.0.

Meta has released Muse Spark, a proprietary AI model with restricted access via API and portal invite only—a striking reversal from CEO Mark Zuckerberg's 2024 manifesto championing open-source AI. The model claims performance matching top competitors from OpenAI, Anthropic, and Google, trained with an order of magnitude less compute than Llama 4.

YouTube is rolling out an AI avatar feature that lets users create photorealistic versions of themselves for YouTube Shorts. Users record a live selfie and voice prompts to generate an avatar that can create up to 8-second video clips. The feature includes watermarks, digital labels (SynthID and C2PA), and AI-generated content disclosures.

Google is introducing 'notebooks' to the Gemini app, a new organizational feature that lets users create personal knowledge bases across chats and files. The notebooks sync directly with NotebookLM and are rolling out first to Google AI Plus, Pro, and Ultra subscribers on web, with mobile and free user access coming in the following weeks.

Meta has released Muse Spark, the first model from Meta Superintelligence Labs, a unit assembled under chief AI officer Alexandr Wang following Meta's $14.3 billion investment in Scale AI. The natively multimodal model features a "Contemplating" reasoning mode with parallel sub-agents and marks Meta's break from its open-source Llama heritage by operating as closed source.