Meta has unveiled its Muse Wonder and Muse Wonder Full AI models, targeting coding, reasoning, and complex professional tasks. Both models include Active Memory for tracking conversations and Active Focus for error recovery. Muse Wonder Full is restricted to paid Muse Code subscribers, with plans ranging from $50 to $200 per month. The models also feature safeguards against sharing private information and can refuse harmful requests, supporting use cases in law, finance, health, and defensive cybersecurity.
CEO Mark Zuckerberg rejected calls to slow AI development, declaring that Meta has caught up with OpenAI and Anthropic using Muse Wonder. He highlighted the model's ability to handle complex tasks like reviewing lease agreements and analyzing financial or health data. Meta plans to integrate Muse Wonder into its Muse AI assistant app, and the model is available to developers through Meta's Model API and Muse Code.
Meta shares rose nearly 3% following the launch. Analysts at Wells Fargo ranked Muse Wonder Full among the most advanced AI models, behind only Anthropic's Claude Opus 5.5. The launch positions Meta to attract more paying users and grow revenue beyond advertising. Meanwhile, DXC Technology partnered with LangChain to scale agentic AI across 125 customer environments through its OASIS platform, cutting filesystem incident resolution from 53-90 minutes to 3-6 minutes and complex problem management from 2-6 days to 8-10 minutes.
In other enterprise AI news, Business Core Solutions won an operations and efficiency award for its Symphony platform, which combines rule-based orchestration, conversational AI through Maestro, and autonomous agentic AI through isAI. Farsight deployed its AI platform at Deutsche Bank, letting bankers create pitch decks, financial models, and memos while keeping all data inside the bank's own infrastructure. NYU Stern launched Viva, a voice AI system for oral exams that uses three agents and a council of three LLMs to score and grade student performance.
Trust remains a key concern for workplace AI deployment. Expert Grant argues that a written value case and pre-deployment baselines on three metrics are required before any agent runs, with ownership split across integration leads, risk and compliance teams, and internal audit. South Korea approved AI law amendments requiring government training data to be linked through the national AI Hub platform. CIOs from Australian superannuation funds warned that the AI-driven market rally could face a sharp correction, though they still expect AI to deliver productivity gains over time.
Key Takeaways
- Meta launched Muse Wonder and Muse Wonder Full AI models with Active Memory and Active Focus features, available through Meta Model API and Muse Code at $50-$200/month.
- Zuckerberg rejects slowing AI development, says Meta has caught up with OpenAI and Anthropic using Muse Wonder.
- Wells Fargo ranks Muse Wonder Full behind only Anthropic's Claude Opus 5.5 among the most advanced AI models.
- DXC and LangChain cut incident resolution time from 53-90 minutes to 3-6 minutes using agentic AI in OASIS across 125 customer environments.
- Business Core Solutions wins AI award for Symphony platform with three modes: rule-based orchestration, conversational AI (Maestro), and autonomous agentic AI (isAI).
- Farsight deploys AI at Deutsche Bank, keeping all data within the bank's infrastructure for pitch decks, models, and memos.
- NYU Stern launches Viva voice AI for oral exams, using three agents and a council of three LLMs to grade student performance.
- Expert Grant requires written value cases and pre-deployment baselines before deploying AI agents in workplaces, with three-line ownership model.
- South Korea approves AI law amendments linking government training data to the national AI Hub platform.
- Australian CIOs warn AI market rally could face sharp correction but still expect long-term productivity gains.
Meta launches Muse Wonder and Muse Wonder Full AI models
Meta released two new premium AI models called Muse Wonder and Muse Wonder Full. They are built for coding, reasoning and complex professional tasks. The models include Active Memory to track conversations and Active Focus to recover from errors. Muse Wonder Full is available only to paid Muse Code subscribers with plans from $50 to $200 per month.
Meta outlines Muse Wonder safeguards for AI agents
Meta detailed safety features for Muse Wonder and Muse Wonder Full, which support writing, reasoning, coding and research. The models avoid sharing private information without permission and can refuse harmful requests. They support defensive cybersecurity and professional work in law, finance and health. Muse Wonder is available through the Meta Model API and Muse Code, while Muse Wonder Full is limited to Muse Code subscribers.
Meta CEO Zuckerberg rejects calls to slow AI development
Mark Zuckerberg said Meta has caught up with OpenAI and Anthropic using the Muse Wonder model. He announced further investment in AI and rejected calls to slow development over safety concerns. Muse Wonder can handle complex tasks like reviewing lease agreements and analyzing financial or health data. Meta also plans to add the model to its Muse AI assistant app.
Meta shares rise as new AI model family impresses investors
Meta shares rose nearly 3% after launching the Muse Wonder model family. Analysts including Wells Fargo said Muse Wonder Full ranks among the most advanced AI models, behind only Anthropic's Claude Opus 5.5. Muse Wonder is available to developers through Meta's Model API and Muse Code. The launch may help Meta attract more paying users and grow revenue beyond advertising.
DXC and LangChain advance agentic AI for IT operations
DXC Technology partnered with LangChain to improve agentic AI for critical IT operations. The work uses LangGraph and LangSmith inside DXC OASIS, which is scaling toward 125 customer environments. OASIS agents have cut filesystem incident resolution from 53-90 minutes to 3-6 minutes and complex problem management from 2-6 days to 8-10 minutes. The partnership focuses on governance, human oversight and enterprise context for AI agents.
Business Core Solutions wins AI award for Symphony platform
Business Core Solutions won the AI in Operations and Efficiency award at the TOI AI Quotient Awards 2026 in Mumbai on September 10, 2026. Its Symphony platform uses three modes including rule-based orchestration, conversational AI through Maestro, and autonomous agentic AI through isAI. The platform helps enterprises automate workflows across ERP systems, cloud environments, and IT operations. BCS operates globally across the United States, Europe, the Middle East, Australia, and India.
Trust is the real blocker for AI agents at work
An expert named Grant argues that trust must be engineered before deploying AI agents in the workplace. She requires a written value case and a pre-deployment baseline on three metrics before any agent can run. Ownership follows a three-line model where the integration lead owns output, risk and compliance challenge it, and internal audit decides on material risk. Evidence for autonomy comes from an auditable decision log, deterministic reconciliation code, and a model inventory record.
KT launches eight Google AI subscription products
Korea Telecom launched eight subscription products that combine Google AI services with coffee and lifestyle benefits. The offering includes three standalone AI subscriptions and five bundles pairing AI services with consumer perks. Standalone options include Google AI Plus with 400 GB or 2 TB of storage and Google AI Pro with 5 TB. The products were reported by Asia Today citing company sources.
Stout acquires Flightline to advance AI strategy
Stout acquired Flightline to build agentic AI into the way its professionals work. The technology has been developed inside real client workflows to test how AI can augment team work. The goal is to create more capacity for Stout professionals to focus on questions requiring experience, judgment, and client interaction. The acquisition aims to bring greater depth of analysis and insight to client work.
NYU Stern uses Voice AI for oral assessments
NYU Stern's Schar School developed Viva by NYU Stern to conduct and grade oral exams using voice AI. The system uses three agents for authentication, project discussion, and case discussion to prevent conversational drift and isolate failures. A council of three LLMs independently scores transcripts and revises after seeing each other's reasoning. The platform was built using ElevenLabs Conversational AI and addresses gaps between written student work and actual understanding.
Farsight deploys secure AI tools at Deutsche Bank
Farsight has deployed its AI platform at Deutsche Bank, allowing bankers to use AI with internal data while keeping information inside the bank's technology environment. The platform supports tasks like creating pitch decks, financial models, and internal memos through PowerPoint, Excel, Word, and email. It operates within the bank's own infrastructure, preserving proprietary knowledge and client confidentiality. Deutsche Bank managing director George Tierney said the combination of control and customization is critical for AI in a regulated enterprise.
SalesCloser launches AI onboarding platform Sky
SalesCloser Technologies CEO Ali Tajskandar launched Sky, an AI-powered onboarding platform that helps businesses build customized AI sales agents. The company has filed ten patent applications and received three U.S. patents in conversational AI sales automation. SalesCloser serves customers mainly in technology and financial services, with the U.S. as its largest market, and operates at roughly 70% gross margins. It also announced a collaboration with Tendril to combine AI with human-assisted sales.
South Korea approves AI law amendments
South Korea's cabinet approved amendments to the AI Basic Act and Data Industry Act to improve data management and availability for AI development. The changes require AI training data from government projects to be linked to and provided through the national AI Hub platform. This establishes a framework for managing and sharing government-generated datasets. The goal is to expand access to training data while improving oversight.
CIOs warn AI boom could face sharp correction
Chief investment officers from major Australian superannuation and wealth management funds warned that the AI-driven market rally could face a sharp and vicious correction. Russell Clarke of VFMC said the AI juggernaut is inflating equity and private markets, and a disappointment in earnings or capex could trigger serious danger. JANA's Matt Wacher called it an earnings bubble, while MLC Super's Dan Farmer noted growing downside risk in equities. Despite the warnings, all three said they still see AI generating productivity gains over time.
Sources
- Meta launches Muse Wonder and Muse Wonder Full AI models: Features, pricing and availability
- Meta outlines Muse Wonder safeguards for AI agents acting on your behalf
- Meta AI: Zuckerberg Rejects Calls To Slow Development, Backs More Investment
- Meta shares rise for a second straight day as new AI model family impresses
- DXC and LangChain Advance Agentic AI Harnesses for Mission-Critical IT Operations
- Business Core Solutions Wins AI in Operations & Efficiency Award at TOI AI Quotient Awards 2026
- Trust Is the Real Blocker for Agents at Work
- KT launches eight Google AI subscription products with lifestyle benefits
- Stout Acquires Flightline to Advance Its AI Strategy
- Scalable and Personalized Oral Assessments Using Voice AI
- Farsight brings secure AI workflows to Deutsche Bank
- SalesCloser launches AI onboarding platform Sky, highlights patent growth and margin expansion potential
- Korea approves AI law amendments to expand access to training data
- CIOs warn that ‘AI juggernaut’ could face ‘vicious’ correction
Comments
Please log in to post a comment.