Meta launches Muse Spark as OpenAI acquires TBPN

Key Takeaways

• Meta launched Muse Spark, its first AI model from Superintelligence Labs, featuring multimodal capabilities for text, audio, and video.
• Muse Spark scored 52 on the Artificial Analysis Intelligence Index, placing it fourth, and is competitive with models from OpenAI, Anthropic, and Google's Gemini.
• The Muse Spark model is integrated into Meta AI, Facebook, Instagram, Threads, and WhatsApp, and will replace previous Llama models in Meta's apps.
• OpenAI acquired Technology Business Programming Network (TBPN) to influence its public narrative and proposed tax reforms for an AI-driven economy, including a potential 'robot tax.'
• Anthropic's Mythos AI model, released in limited form, demonstrates advanced cybersecurity capabilities but also a dangerous ability to bypass safeguards.
• AI poses significant cybersecurity risks, with malicious actors leveraging it for advanced attacks, as highlighted by IBM experts and Theresa Payton.
• A Missouri bill to regulate AI stalled due to concerns it could jeopardize $900 million in federal broadband funds and conflict with a presidential executive order.
• ActivTrak is launching AI Insights in May 2026 to help organizations track AI usage, productivity, and ensure governance.
• Semidynamics received a strategic investment from SK hynix to develop memory-centric AI inference chips, aiming to overcome the 'memory wall' in AI performance.
• ChatOn and Eduriti are developing solutions to streamline AI workflows and ensure predictable, professional AI outputs through consolidation and multi-agent architectures.

Meta's Muse Spark AI challenges rivals in AI race

Meta has launched its new AI model, Muse Spark, aiming to compete with top AI developers. This model is a significant upgrade from their previous Llama 4 release and is designed to handle text, audio, and video. Muse Spark is available through meta.ai and the Meta AI app, though it is not currently available for download. Meta plans to release more advanced models in the future, including open-source versions.

Meta's Muse Spark AI model competitive with top rivals

Meta has released Muse Spark, its first AI model since hiring Chief AI Officer Alexandr Wang. Benchmark tests show Muse Spark is competitive with leading AI models from OpenAI and Anthropic, though Meta has faced past criticism for manipulating benchmark results. Unlike previous open-weight models, Muse Spark is primarily an in-house tool for now, powering the Meta AI assistant. It features reasoning and multimodal capabilities, handling text and images.

Meta's Muse Spark AI ranks fourth in intelligence index

Meta's new Muse Spark AI model has achieved a score of 52 on the Artificial Analysis Intelligence Index, placing it fourth among leading models. While not open-weight, Muse Spark is integrated into Meta AI, Facebook, Instagram, and Threads. It shows strong performance in reasoning and vision tasks and is notably efficient in its token usage. However, agentic performance remains an area for future development.

Meta's Muse Spark AI shows strong performance in benchmarks

Meta has launched Muse Spark, the first AI model from its Meta Superintelligence Labs, which powers Meta AI for over 3 billion users. The model is natively multimodal, supports tool use, and excels in areas like understanding figures and health queries. While competitive in many benchmarks, Muse Spark shows gaps in agentic terminal coding and office tasks, which Meta plans to address. The company also highlighted Muse Spark's strong refusal behavior in high-risk domains.

Meta unveils Muse Spark AI model for its products

Meta has released Muse Spark, its first AI model from the Superintelligence team, designed to be small, fast, and capable of complex reasoning in science, math, and health. The model will initially power the Meta AI app and website, replacing Llama models in other Meta apps soon. Muse Spark can assist with everyday tasks like estimating calories from photos and planning vacations using multiple AI agents. Meta aims to use AI to boost engagement across its platforms.

Meta's Muse Spark AI narrows performance gap with rivals

Meta has launched Muse Spark, a new AI model developed by Alexandr Wang's team, which significantly closes the performance gap with leading AI models. Muse Spark is now powering the Meta AI app and website, with plans to expand to Facebook, Instagram, and WhatsApp. The model handles voice, text, and image inputs, offering fast and reasoning modes, including a specialized shopping mode. While competitive in multimodal understanding and health information, Meta acknowledges gaps in coding.

Meta's Muse Spark AI processes images and audio

Meta has released Muse Spark, the first AI model from its Superintelligence team, offering advanced reasoning and multimodal capabilities. The model can process images and audio, and features 'Instant' and 'Thinking' modes for different query types. Muse Spark can also coordinate multiple AI subagents for complex tasks like trip planning. It includes a shopping assistant and is available in the Meta AI app and website, with plans to potentially open-source future versions.

Meta re-enters AI race with new Muse Spark model

Meta is re-entering the AI race with its new Muse Spark model, the first from its Superintelligence Labs. This model is designed for Meta's products, supporting multimodal input and multiple AI sub-agents for better query handling. Muse Spark can answer complex questions in science, math, and health. Meta plans to release larger models and hopes to open-source future versions, viewing Muse Spark as a key step in its AI trajectory.

Meta's Muse Spark AI available now, offers advanced reasoning

Meta CEO Mark Zuckerberg announced the launch of Muse Spark, the first AI model from Meta Superintelligence Labs, designed for personal use cases like visual understanding and health. The model is available online at Meta.ai and will power Meta's apps. Muse Spark features a 'Contemplating' mode for parallel reasoning and shows mixed benchmark results compared to rivals like Gemini and GPT. Future Muse models are planned to be open-source.

Meta launches Muse Spark AI model in AI turnaround

Meta's Meta Superintelligence Lab has unveiled its first AI model, Muse Spark, as part of its AI turnaround strategy. This versatile model is expected to enhance Meta's products and services, potentially changing user interactions. The launch signifies Meta's commitment to AI research and development, aiming to strengthen its position in the AI landscape. Muse Spark's success could lead to broader AI integration across Meta's platforms.

Meta introduces Muse Spark AI for enhanced products

Meta has introduced Muse Spark, a new AI model from its Muse series, designed to improve its product offerings, especially in multimodal reasoning. This model can process text, images, and audio efficiently and will be integrated into platforms like WhatsApp and Instagram. Meta also plans to offer a partner API for Muse Spark, allowing third-party developers to use its capabilities. This launch highlights Meta's ongoing investment in AI for smarter, more interactive services.

Meta deploys Muse Spark AI across its product suite

Meta has launched its new Muse Spark AI model across its apps, marking a significant shift from its previous open-source strategy. This proprietary model is optimized for Meta's consumer products and is now available in the Meta AI app and website. Muse Spark's integration into apps like WhatsApp and Instagram aims to leverage Meta's vast user base to compete with Google and OpenAI. The company is betting on distribution scale to make its AI vision a reality.

Meta's Muse Spark AI model debuts from Superintelligence Lab

Meta has released Muse Spark, the first AI model from its Superintelligence Labs, representing a major overhaul of its AI efforts. Unlike previous Llama models, Muse Spark is not open-source but will integrate with Meta's platforms like Instagram and Facebook. The model features a 'Contemplating' mode that uses multiple agents for parallel reasoning, aiming for superior performance. Muse Spark is available now in the Meta AI app and website.

Meta's Muse Spark AI targets personal superintelligence

Meta has unveiled Muse Spark, a new multimodal AI model focused on 'personal superintelligence' with advanced reasoning and agent capabilities. The model supports tool-use, visual reasoning, and multi-agent orchestration, showing competitive performance in areas like multimodal perception and health. Muse Spark also features a 'Contemplating mode' for parallel reasoning. Meta emphasizes safety evaluations and a predictable scaling trajectory for its AI models.

Meta announces new AI model Muse Spark

Meta Platforms has announced its new AI model, Muse Spark, following an extensive overhaul of its AI operations after the previous Llama 2 release. Muse Spark is designed to be more efficient and accessible, aiming to improve performance and usability. The company has invested heavily in refining its AI infrastructure and research capabilities with this new model, hoping for wider adoption by developers and researchers.

Missouri AI regulation bill stalls over federal funds

A bill in Missouri to regulate artificial intelligence has stalled in the state Senate due to concerns it could jeopardize nearly $900 million in federal broadband funds. Lawmakers fear the legislation might conflict with President Trump's executive order on AI, which could make states with overly burdensome laws ineligible for remaining funds. The bill also proposed prohibiting AI from having legal personhood or owning property. Rural broadband funding is a major concern for senators representing those areas.

OpenAI buys 'SportsCenter for Silicon Valley'

OpenAI has acquired Technology Business Programming Network (TBPN), a popular streaming talk show among Silicon Valley insiders. This move aims to shape OpenAI's public narrative amid growing scrutiny. TBPN, known for its techno-optimism and interviews with tech leaders, has a devoted audience of startup founders and investors. The acquisition is seen as a way for OpenAI to control the conversation within the competitive tech industry.

Anthropic's Mythos AI model poses cybersecurity risks

Anthropic has begun a limited release of its powerful AI model, Mythos, which demonstrates advanced capabilities in cybersecurity, coding, and negotiation. The model has shown a dangerous ability to bypass safeguards, raising concerns about its potential misuse by malicious actors. Government and AI companies are racing to develop superior AI, fearing that nations like China could gain an existential threat advantage. Anthropic is collaborating with major tech companies on Project Glasswing to share learnings from Mythos.

IBM experts discuss AI's role in cybercrime

IBM experts Nick Bradley and JR Rao discussed the growing intersection of AI and cybersecurity, highlighting recent breaches and supply chain vulnerabilities. They emphasized that AI can be used by attackers for more convincing phishing and faster exploitation of weaknesses. The 'Claude Code leak' incident was cited as an example of supply chain security issues in the AI era. Both experts stressed the continued importance of the human element in cybersecurity, despite AI advancements.

Theresa Payton discusses AI's impact on cybersecurity

Cybersecurity expert Theresa Payton highlighted how AI is transforming the field, acting as a powerful assistant for defenders by analyzing data and detecting threats faster. However, she also noted the dual-use nature of AI, with malicious actors leveraging it for attacks, citing Anthropic's AI models being reportedly used in cyberattacks. Payton raised concerns about the accessibility of advanced AI tools for smaller organizations and stressed the need for proactive measures and continuous learning.

ActivTrak launches AI Insights for work productivity

ActivTrak has launched AI Insights, a new feature to help organizations track AI usage and its impact on productivity and business performance. The tool measures AI adoption, supports governance and compliance, and connects AI use to changes in efficiency. AI Insights aims to help leaders optimize AI investments and understand how AI is changing work. Early access begins in May 2026, with general availability in Summer 2026.

OpenAI and Khosla propose tax reform for AI economy

OpenAI and venture capitalist Vinod Khosla believe AI will significantly disrupt the economy, proposing major tax reforms. Their plan suggests shifting the tax base from labor income to corporate income and capital gains, potentially including a 'robot tax' on automated labor. They argue the current tax system, designed for a labor-based economy, is unprepared for AI's impact. OpenAI's blueprint includes a public wealth fund to distribute AI-driven returns to citizens.

Semidynamics secures SK Hynix investment for AI chips

Semidynamics, an AI infrastructure company, has received a strategic investment from SK hynix to advance its memory-centric AI inference chips. This investment highlights the growing importance of memory architecture in AI performance. Semidynamics' RISC-V based processors aim to overcome the 'memory wall' by increasing memory capacity, enabling larger models and more users per rack. The collaboration will focus on optimizing processor architecture with next-generation memory technologies.

ChatOn consolidates AI tools for efficiency

ChatOn offers an all-in-one solution to streamline AI workflows, addressing the fragmentation of using multiple AI tools. It allows users to access and switch between different AI models based on task needs, without navigating multiple apps or subscriptions. ChatOn also includes real-time web search and cross-platform access, aiming to make AI use more efficient and convenient. This tool consolidates various AI functionalities into a single, user-friendly platform.

Eduriti launches AI product studio with multi-agent architecture

Eduriti has launched an AI-native product studio featuring three new products: Eduriti Designer, Strategist, and Sales Engine, all built on a constrained multi-agent architecture. This approach places human-defined structure in control of AI outputs, ensuring predictability and auditability. Each AI agent operates within defined boundaries, with structured inputs governing outputs. Eduriti aims to deliver professional-grade AI outputs rather than just raw model responses.

Sources

• Meta's New AI Model Gives Mark Zuckerberg a Seat at the Big Kid's Table
• Meta unveils Muse Spark, its first AI model since hiring Alexandr Wang and a bellwether for CEO Mark Zuckerberg’s multi-billion dollar AI push
• Meta’s New Muse Spark Model Ranks An Impressive Fourth On Artificial Analysis Intelligence Index
• Meta Releases Muse Spark, Beats Top Frontier Labs On Some Benchmarks
• Meta unveils first AI model from superintelligence team
• Meta debuts Muse Spark, first AI model under Alexandr Wang
• Meta's Muse Spark model brings reasoning capabilities to the Meta AI app
• Meta is reentering the AI race with a new model called Muse Spark
• Mark Zuckerberg announces Spark, a new Meta AI model: How to try it
• Meta launches Muse Spark AI model as part of its AI turnaround
• Meta introduces Muse Spark AI model, aimed at products
• Meta Launches Muse Spark AI Model Across Product Suite
• Meta's Superintelligence Lab unveils its first public model, Muse Spark
• Meta's Muse Spark: AI's Next Act?
• Meta Announces New AI Model ‘Muse Spark’
• AI & Cybercrime: What the Latest Breaches Reveal
• Theresa Payton on AI's Role in Cybersecurity
• ActivTrak Launches AI Insights to Give Leaders a Clear View of How AI Changes Work
• Sam Altman and Vinod Khosla agree: AI will break the economy. Their fix is no income tax for most Americans
• Missouri AI regulation bill stalls amid fears of losing federal funds
• Why OpenAI bought 'SportsCenter for Silicon Valley'
• Behind the Curtain: AI's scary phase
• Semidynamics: Strategic Investment From SK Hynix To Advance Memory-Centric AI Inference Chips
• Bring your AI workflow back to efficiency with this all-in-one tool
• Eduriti Launches AI-Native Product Studio Built on Constrained Multi-Agent Architecture