Modelyo, an Israeli startup, raised $17 million led by OlivTree to build its confidential cloud platform. The technology helps organizations use AI and cloud services while keeping data and encryption keys protected. It relies on confidential computing and zero-trust architecture to secure information during processing, with a focus on governments, banks, and defense organizations.
NVIDIA introduced the OpenShell Reference Architecture, a framework designed to secure and govern AI agents within enterprise systems. The design uses hardware-based security including NVIDIA Vera CPUs and BlueField-4 data processing units, along with an open-source secure runtime that isolates agent workloads. It supports cloud, on-premises, hybrid, and edge environments, and NVIDIA reports that multiple technology providers are already integrating with the system.
VDR.ai and LLM.co partnered to offer custom AI hardware and private language models tailored for financial institutions. The joint solution targets investment banks, private equity firms, and asset managers, allowing them to deploy private AI systems on-premises or in isolated cloud environments. The platform supports document analysis, financial data extraction, automated reporting, and tasks like mergers and acquisitions due diligence.
Carolyn Duby from Cloudera highlighted the risks of AI agents that can act autonomously, urging organizations to extend zero-trust security and continuously authorize agent actions. Separately, a survey found that 62 percent of respondents deployed AI tools before setting up safeguards, and three in ten would let an AI agent run a security investigation or change a production environment without human oversight.
Anthropic's Deputy Chief Information Security Officer, Clinton, warned that cybersecurity threats will worsen before they improve, with a vulnerability tsunami potentially outpacing critical infrastructure patching. He advised security leaders to prioritize strong defenses now rather than rushing to adopt AI security tools, believing AI scaling laws will eventually shift the advantage to defenders.
Spotify co-CEO Gustav Söderström said the company is preparing for an agentic AI future by building its own AI agent that works with third-party assistants to handle music requests. Spotify has deals with OpenAI and others to integrate services, and the company says it will not share user data with outside agents. Dialogica hired Dylan Gittleman as Director of Business Development for its voice-first AI secretary called Dia, aimed at large law firms.
Microsoft partnered with the Public Library Association to expand free AI literacy training through libraries. The initiative, called Building AI Skills and Agency Through Libraries, provides resources, webinars, and a Community of Practice for library workers. Florida Atlantic University received a $500,000 National Science Foundation grant to build a research and training framework for AI-enabled robotics, creating shared cyberinfrastructure with digital twins and robotic testbeds.
Key Takeaways
- Modelyo raised $17 million led by OlivTree to develop its confidential cloud platform for protecting AI data
- NVIDIA launched OpenShell Reference Architecture using Vera CPUs and BlueField-4 DPUs to secure AI agents
- VDR.ai and LLM.co partnered to deliver private AI infrastructure for financial institutions including investment banks and asset managers
- 62 percent of organizations deployed AI tools before establishing safeguards, and 75 percent have faced at least one AI agent problem
- Anthropic's CISO warned of a coming vulnerability tsunami and urged stronger defenses before rushing to adopt AI security tools
- Spotify built its own AI agent and has deals with OpenAI while stating it will not share user data with outside agents
- Dialogica hired Dylan Gittleman to grow sales for its Dia voice-first AI secretary targeting Am Law 100 firms
- Microsoft partnered with the Public Library Association to expand free AI literacy training through libraries
- Florida Atlantic University received a $500,000 NSF grant for AI robotics workforce training using digital twins and testbeds
- Cloudera's Carolyn Duby recommended assigning AI agent permissions by individual tasks rather than broad access
Modelyo raises $17M to protect AI data in the cloud
Israeli startup Modelyo raised $17 million led by OlivTree to develop its confidential cloud platform. The technology lets organizations use AI and cloud services while keeping data and encryption keys secure. It uses confidential computing and zero-trust architecture to protect data during processing. The company targets governments, banks, and defense organizations. Modelyo has 30 employees and was founded in 2024.
VDR.ai and LLM.co partner on private AI for finance
VDR.ai partnered with LLM.co to offer custom AI hardware and private language models for financial institutions. The partnership targets investment banks, private equity firms, and asset managers. Organizations can deploy private AI systems on-premises or in isolated cloud environments. The solution supports document analysis, financial data extraction, and automated reporting. It aims to help institutions use AI while keeping sensitive data controlled.
VDR.ai and LLM.co deliver private AI to financial firms
VDR.ai and LLM.co announced a partnership combining AI document management with private AI infrastructure. The joint offering serves financial organizations handling confidential documents. It uses custom GPU workstations and dedicated computing clusters for private large language model deployments. The system supports tasks like mergers and acquisitions due diligence and credit underwriting. Installations can operate on-premises or in air-gapped configurations.
NVIDIA launches OpenShell to secure AI agents
NVIDIA introduced the OpenShell Reference Architecture to secure and govern AI agents in enterprise systems. The framework uses hardware-based security including NVIDIA Vera CPUs and BlueField-4 data processing units. It includes an open-source secure runtime that isolates agent workloads and monitors access to files and networks. The architecture supports cloud, on-premises, hybrid, and edge environments. NVIDIA says multiple technology providers are integrating with the system.
Cloudera expert on governing AI agent security
Carolyn Duby from Cloudera discussed AI-agent security in a Q&A interview. She explained that organizations must consider the blast radius of AI agents that can act autonomously. She recommended extending zero-trust security to continuously authorize agent actions. Her advice includes assigning permissions by individual tasks rather than broad access. A Cloudera survey found 46% of Asia-Pacific respondents cited data security as the main driver of AI infrastructure change.
Dialogica Hires Dylan Gittleman to Grow Legal AI Sales
Dialogica added Dylan Gittleman as Director of Business Development for its voice-first AI secretary called Dia. Dia is designed for large law firms and aims to reduce non-billable work while keeping lawyer judgment and client confidentiality. Gittleman has experience in enterprise sales, investment banking, and legal technology. He will focus on reaching Am Law 100 firms and Fortune 500 legal departments.
Microsoft and ALA Expand AI Literacy Through Libraries
Microsoft partnered with the Public Library Association, a division of the American Library Association, to expand free AI literacy training. The initiative called Building AI Skills and Agency Through Libraries will provide resources, webinars, and a Community of Practice for library workers. It aims to help people learn practical AI skills and understand risks like bias and safety. This effort supports Microsoft's broader Community-First AI Infrastructure and Microsoft Elevate commitments.
FAU Receives NSF Grant for AI Robotics Workforce Training
Florida Atlantic University received a $500,000 National Science Foundation grant to build a research and training framework for AI-enabled robotics. Led by Xiangnan Zhong, the project will create shared cyberinfrastructure with digital twins and robotic testbeds. Over four years, it aims to help students and researchers run realistic experiments across virtual and physical environments. The goal is to prepare a skilled workforce for AI robotics and cyber-physical systems.
Anthropic CISO Warns Cyber Threats Will Worsen
Clinton, Anthropic's Deputy Chief Information Security Officer, warned that cybersecurity threats will get worse before they improve. He said a vulnerability tsunami is coming and critical infrastructure may not be patched in time. He advised security leaders to focus on strong defenses now rather than rushing to adopt AI security tools. He believes AI scaling laws will eventually shift the advantage to defenders.
Spotify Says It Is Ready for Agentic AI Future
Spotify co-CEO Gustav Söderström said the company is well positioned for agentic AI. Spotify built its own AI agent that works with third-party assistants to handle music requests while keeping direct consumer relationships. The company has deals with OpenAI and others to integrate its services. Spotify says it will not share user data with outside agents but will learn more about listener interests through agent interactions.
Micron stock climbs as AI memory plans take focus
Micron Technology is a memory-chip producer. Its stock rose as the company highlighted its AI memory roadmap. President and technology chief Scott DeBoer discussed the plan. Investors expect strong execution. The stock trades at a premium to its GF Value estimate of $793.59. Investors should watch HBM production growth, gross margins and capital spending.
AI agent permissions outpace security safeguards
Most organizations have given AI agents authority without proper controls. Three in four organizations have faced at least one problem with their AI agents. Three in ten would let an AI agent run a security investigation or change a production environment without human oversight. Sixty-two percent of respondents say they deployed AI tools before setting up safeguards. Experts recommend starting with the agent that can access the most sensitive data and limiting its actions.
Clover School District blocks all AI programs except one
The Clover School District is limiting AI use on student-issued laptops. Only one program called SchoolAI will be allowed. This restriction will last through spring 2027. The district decided to block other AI programs to manage student safety and learning.
Sources
- Modelyo raises $17 million to let companies use AI without exposing their secrets
- VDR.ai Partners with LLM.co to Deliver Custom AI Hardware and Private LLM Deployments for Financial Institutions
- VDR.ai Partners with LLM.co to Deliver Custom AI Hardware and Private LLM Deployments for Financial InstitutionsÂ
- NVIDIA's OpenShell architecture to govern AI agents
- Carolyn Duby on AI-agent security and zero trust
- Dialogica Adds Dylan Gittleman To Founding Sales Team To Expand Legal AI Adoption
- Microsoft and the American Library Association expand AI literacy
- FAU Receives NSF Grant to Prepare Future AI, Robotics Workforce
- Things Will Get Worse Before They Get Better: Anthropic Deputy Chief Information Security Officer On Recent Cyber Incidents
- Spotify Claims It Is 'Unusually Well-Positioned' for an Agentic AI Takeover
- Micron Stocks Rise as AI-Memory Roadmap Takes Center Stage
- AI agent permissions are outrunning security
- Clover School District will temporarily block the use of all AI programs but one. Here’s why.
Comments
Please log in to post a comment.