Meta Launches Muse Wonder Models
Meta is introducing the Muse Wonder series with Muse Wonder and Muse Wonder Full, our most capable models to date, according to Meta. The models deliver state-of-the-art performance on personal and professional agentic tasks. They bring capability to expert domains including law, finance, and health, along with frontier coding performance for people building everything from a game to an app for a new business.
Features Designed for Professional Use
Muse Wonder and Muse Wonder Full are equipped with Active Memory, the ability to carry what matters to you across conversations and bring back the details that are relevant to the task at hand. They’re also equipped with Active Focus, the ability to keep working toward your goal when a task gets hard, try other approaches, recover from errors, and follow through, asking for clarification when needed and checking in before following through on anything important. The models were trained to avoid sharing private information without permission, turn down requests to assist in performing actually harmful acts, and work to improve security and strengthen defenses against cyber attacks. Developers and security teams can use models like Muse Wonder to find flaws in their own code and strengthen their systems.
Built on Safety and Scaling Framework
Meta trains, evaluates, and releases its most capable AI models in line with the standards set out in its Meta Superintelligence Scaling Framework. A full report with additional safety evaluations will be available soon. This approach ensures the models meet rigorous benchmarks for reliability and risk mitigation before release. The framework guides how Meta develops frontier models while addressing concerns about misuse and unintended consequences in high-stakes professional environments.
Strategic Shift Toward AI Partnerships
The launch of Muse Wonder comes as Meta shifts from solo AI development to strategic collaborations, aiming to strengthen its position against Google and Microsoft in the enterprise AI race. This pivot contrasts with recent controversial scraping terms for news content, which allow Meta to use web scrapers to crawl and collect content from websites ‘associated with’ business accounts on its platforms. The update, set to go into effect on October 30, will include news publishers with business accounts. Meta may use any content from such domains to provide, develop, improve, research, train, and maintain the integrity of Meta Company Products. Currently, there is no option for business accounts to opt out of these new terms ahead of October 30. Meta declined to answer a list of questions about these updated terms. Meta’s decision to tether access to its business products to broad scraping permissions echoes a concept in antitrust law called ‘tying’.
Implications Of Meta’s Scraping Terms For News Publishers
According to Matt Topic, a media and intellectual property attorney at Loevy & Loevy, the new scraping terms are a “remarkable provision” that could give Meta access to material worth millions of dollars in a licensing market. James Grimmelmann, a professor of digital and information law at Cornell Law School, notes that these terms represent permission for Meta to train AI on any content from news publishers’ sites, potentially allowing Meta to deploy that content in competition with the publishers themselves.
Currently, there is no option for business accounts to opt out of these new terms ahead of October 30, leaving publishers with the choice to either trust that Meta doesn’t intend to use the full extent of these terms or stop using business accounts altogether, as stated by Grimmelmann. Meta declined to answer questions about these updated terms, with a public affairs spokesperson saying that the new terms consolidate existing language and will not change how Meta crawls publishers in the future.
However, legal experts point out that any attempt to train Meta’s AI products with content collected under these new terms would still hinge on unsettled legal questions about AI training and fair use, with several lawsuits between publishers and AI developers currently ongoing. Benjamin Sobel, a law professor at the University of Wisconsin, notes that Meta’s decision to tie access to its business products to broad scraping permissions echoes the concept of “tying” in antitrust law.
The Meta Superintelligence Scaling Framework
Meta’s development of frontier models like Muse Wonder is guided by the Meta Superintelligence Scaling Framework. This framework establishes the rigorous standards for training, evaluating, and releasing the company’s most capable AI models. It specifically addresses the complex challenges of mitigating risks and preventing misuse, particularly in high-stakes professional environments where these models might be deployed.
The framework ensures models undergo extensive safety evaluations before public release. A full report detailing these evaluations is scheduled to be released soon, providing transparency into the process. Meta states this approach guarantees models meet stringent benchmarks for reliability and risk mitigation, aiming to build trust in their deployment for critical tasks like legal analysis, financial planning, and healthcare assistance, while also strengthening security defenses against cyber threats.
Competitive Positioning Against Google Gemini and Microsoft Copilot
Meta positions the Muse Wonder series as a direct response to competing enterprise AI offerings from Google Gemini and Microsoft Copilot, emphasizing its focus on professional agentic tasks in law, finance, and health. According to Meta, the models deliver state-of-the-art performance on personal and professional agentic tasks, bringing capability to expert domains including law, finance, and health, along with frontier coding performance for people building everything from a game to an app for a new business. This launch reflects Meta’s strategic shift from solo AI development to partnerships aimed at strengthening its position against Google and Microsoft in the enterprise AI race.
Sources
- Meta
- Introducing Muse Wonder and Muse Wonder Full: Frontier Models for Agents
- Meta’s new commercial terms will allow the scraping of news sites, legal experts say
Photo: Exilexi (CC BY 4.0)
