Meta's Muse AI Agent Impacts Users' Privacy

Key Takeaways

• Meta's Muse AI agent can impact users' privacy even when friends adopt it, collecting data on interactions and behavior
• Research from the University of Maryland Baltimore County warns AI in classrooms harms young children's critical thinking development
• The SGLang Summit in San Francisco featured Intel CEO Lip-Bu Tan, OpenAI's Lilian Weng, and sponsors Google, NVIDIA, Meta, and AMD
• OpenAI CEO Sam Altman urged MBA students to build new companies rather than pursue traditional business ideas
• NYU Langone expanded its OpenEvidence partnership to integrate clinical AI hospital-wide for personalized patient care
• Phunware pivots to AI guest intelligence with new products AI Concierge and Itinerary Planner for hospitality and healthcare
• GitHub's 2026 report shows AI agents now make up over 40% of pull requests, creating new security risks
• Identity vendors face challenges managing AI agent ownership when employees depart, risking orphaned access
• HPE and NVIDIA partner to help neoclouds deliver sovereign AI solutions compliant with EU's NIS2 Directive
• GitHub recorded 34.3 billion commits in 2026 as automation grew by more than 220%

How Friends' AI Agent Choices Affect Your Privacy

Privacy law expert Alexis Shore Ingber from the Newhouse School explains how AI agents like Meta's Muse can impact your privacy. A friend's decision to use an AI agent can affect your personal data. AI agents collect data on your interactions, preferences, and behavior. Protecting data requires technical, social, and legal measures.

Research Says AI in Classrooms Hurts Young Kids

Research shows that using AI in classrooms is not a good idea for young children. Amy Tondreau from the University of Maryland Baltimore County says kids lose chances to build critical thinking skills. New York City and Norway have already limited AI use in schools. Experts urge careful, long-term thinking about AI's impact on developing minds.

SGLang Summit Showcases Next-Gen AI Technologies

The inaugural SGLang Summit took place on November 12-13 in San Francisco. The two-day conference covered frontier models, AI agents, robotics, and generative media. Speakers included Intel CEO Lip-Bu Tan, OpenAI researcher Lilian Weng, and Perplexity CEO Aravind Srinivas. Platinum sponsors included Google, Intel, NVIDIA, Meta, AMD, and RadixArk.

Sam Altman Urges MBA Students to Build New Things

OpenAI CEO Sam Altman criticized MBA students for chasing old business ideas instead of new ones. He said AI will perform many tasks faster and cheaper than humans. Altman sees AI disruption as a chance for young professionals to start careers during major change. He believes AI enables companies that would have been impossible to build before.

NYU Langone Expands Clinical AI Across Hospitals

NYU Langone Health expanded its partnership with OpenEvidence to integrate clinical AI into its electronic health record system hospital-wide. The platform uses machine learning to analyze data and help clinicians make better decisions. This integration gives doctors access to real-time data and analytics for more personalized care. The move supports NYU Langone's goal to improve patient outcomes through technology.

Phunware shifts to AI guest intelligence platform

Phunware CEO Dmitry Kroshka announced a major shift from mobile app development to an AI-enabled B2B guest intelligence platform. The company is launching new products called AI Concierge and Itinerary Planner to serve the hospitality and healthcare industries. Phunware plans to expand its sales efforts toward multi-property groups and healthcare systems while using disciplined capital for growth. The company aims to drive long-term shareholder value through active board oversight and selective acquisitions.

GitHub Octoverse 2026 reveals AI security risks

GitHub's State of the Octoverse 2026 report shows that software development is accelerating rapidly with 34.3 billion commits recorded in 2026. AI agents now make up over 40% of pull requests, creating new risks for exposed secrets and credentials. Citizen developers are building more applications using natural language tools, which often requires access to sensitive business systems. Organizations must update their security controls to handle these changes as automation grows by more than 220%.

Identity vendors struggle with AI agent ownership

Major identity security vendors launched AI agent offerings in 2026, but they face a critical problem regarding agent ownership when employees leave. If an engineer who built an agent departs, the agent can become an orphan with standing access or break entirely. Governance platforms like Agentic Fabric focus on managing these departures, while privileged access platforms generate short-lived secrets to reduce risk. Discovery tools like Reco and Token Security help map agents back to human owners to prevent security gaps.

Neoclouds need sovereign AI for operational control

HPE and NVIDIA are partnering to help neoclouds offer sovereign AI solutions for organizations that need strict data control. Enterprises are embedding large language models into their workflows, but public clouds pose risks of data leakage and unpredictable performance. Neoclouds can solve this by building localized AI instances that comply with laws like the European Union's NIS2 Directive. This approach ensures data stays in specific geographic locations and protects against multi-tenant performance issues.

Sources

• Newhouse Faculty Q&A: How Other People’s Choices Shape Your Privacy in the Age of AI Agents
• AI in the classroom? Research finds it’s not a great idea
• Next-Gen AI Hardware, Infrastructure, Frontier Models, and Applications on Center Stage at SGLang Summit
• Sam Altman's "Chronic Disappointment" In MBA Students
• NYU Langone brings OpenEvidence’s clinical AI into the EHR systemwide
• Phunware CEO outlines shift to AI-driven guest intelligence platform with new products and growth strategy.
• GitHub Octoverse 2026: Five Takeaways for Security Teams
• Identity Security Vendors for AI Agents
• The infrastructure playbook for operational sovereignty: Why neoclouds need to offer sovereign AI