The AI industry is rapidly evolving, with a strong focus on agentic AI and security. The RSAC 2026 conference highlighted an 'AI Tragedy of the Commons' due to confusing messages, emphasizing the critical need for protocol-layer security, particularly for agent-to-agent communications. Startups are addressing these concerns, with Noma Security raising $132 million for secure identity and governance. Aembit provides Identity and Access Management for AI agents, while Aurascape launched a Zero-Bypass Model Context Protocol (MCP) Gateway. Geordie AI also won an award for its AI agent security platform, and BlueRock introduced a Trust Context Engine to ensure trust in AI actions.
Hardware integrity is becoming paramount for AI chips, as counterfeit components pose significant risks. A hardware root of trust, embedded at the factory, is crucial for verifying firmware and managing keys, creating an immutable chain of custody to prevent unauthorized code. In a significant move, Nvidia invested $2 billion in Marvell Technology, integrating Marvell's custom XPUs and networking into Nvidia's NVLink Fusion ecosystem. Marvell is a key supplier of custom AI chips for major players like Amazon Web Services (AWS) and Microsoft, and this partnership allows Nvidia to embrace the custom ASIC trend. CoreWeave also demonstrated high performance using NVIDIA Grace Blackwell systems.
Several major AI companies are seeing internal shifts and strategic developments. OpenAI's head of AGI deployment, Fidji Simo, is taking medical leave, with President Greg Brockman overseeing product development. CMO Kate Rouch is also temporarily stepping down for health reasons. Meanwhile, Take-Two reportedly disbanded its AI team, leading to the departure of its head, Luke Dicken, despite CEO Strauss Zelnick's public commitment to generative AI. The broader AI industry, including OpenAI and Google DeepMind, is increasingly focused on automating its own research processes, developing AI systems that can write code and optimize performance.
AI applications are expanding into diverse sectors, from security to education and faith. ZeroEyes has enhanced its AI-based threat detection to include knives and person tracking, integrating with existing security cameras for public safety in various venues. In education, musician Will.i.am is teaching an agentic AI course at Arizona State University, focusing on ethical AI use and preparing adults for technological advancements, even featuring the head of ChatGPT as a guest speaker. Conversely, West Virginia schools are advised against introducing AI and humanoid robots, with critics arguing that such investments are wasteful given existing educational challenges and that humanoid forms offer no practical advantage. A Collibra survey indicates that while leaders plan to increase AI spending, many feel AI is currently under-delivering. The Vatican is also navigating AI's rapid progress, with Catholic chatbots emerging for evangelization, emphasizing responsible innovation and distinguishing human from artificial intelligence.
Key Takeaways
- RSAC 2026 highlighted the 'AI Tragedy of the Commons' and the urgent need for protocol-layer security for AI agents.
- Agentic AI startups are gaining traction, with Noma Security raising $132 million for secure identity and governance solutions.
- A hardware root of trust is essential for AI chip security, preventing counterfeiting and ensuring firmware integrity from manufacturing to deployment.
- Nvidia invested $2 billion in Marvell Technology, integrating Marvell's custom XPUs and networking into its NVLink Fusion ecosystem.
- OpenAI's head of AGI deployment, Fidji Simo, is on medical leave, with President Greg Brockman overseeing product development.
- Take-Two reportedly disbanded its AI team, leading to the departure of its head, Luke Dicken.
- The AI industry, including OpenAI and Google DeepMind, is increasingly automating its own research processes, developing AI systems to write code and optimize performance.
- ZeroEyes expanded its AI threat detection capabilities to include knives and person tracking for enhanced physical security.
- Musician Will.i.am is teaching an agentic AI course at ASU, focusing on ethical AI use and adult preparedness for AI advancements.
- West Virginia schools are advised against introducing AI and humanoid robots due to existing educational challenges and the lack of practical advantages over current technology.
RSAC 2026: AI's 'Tragedy of the Commons' and Agent Security
The RSAC 2026 conference, in its 35th year, highlighted major cybersecurity trends and vendor activities. A key theme was the AI 'Tragedy of the Commons,' where confusing AI messages make it hard for buyers to find real value. The conference also addressed risks in the unsecured protocol layer, particularly with Model Context Protocol (MCP) and agent-to-agent communications. Discussions covered securing agentic landscapes, AI-generated code security blind spots, and future indicators for protocol-layer security. Geordie AI won the 'Most Innovative Startup' award for its AI agent security platform.
Top 12 Agentic AI Startups to Watch in 2026
The AI startup scene is moving incredibly fast, with many new companies focusing on secure identity and governance for AI agents. Startups like Noma Security are investing heavily to build long-term companies, with Noma raising $132 million. Aembit offers Identity and Access Management for Agentic AI, giving AI agents verified identities. Aurascape launched a tool to address security gaps in the Model Context Protocol (MCP) with its Zero-Bypass MCP Gateway. Command Zero provides an agentic-driven platform to automate Security Operations Center tasks.
AI News April 3: BlueRock, Celigo, Cloudera, CoreWeave Updates
This week's AI news includes BlueRock's Trust Context Engine for mapping AI actions and ensuring trust. Celigo introduced Ora and Agent Builder to turn natural language requests into automated tasks. Cloudera is showcasing its 'AI Anywhere' strategy at AWS Summits, focusing on hybrid AI. A Collibra survey reveals leaders plan to increase AI spending, but many feel AI is under-delivering. CoreWeave achieved high performance in MLPerf v6.0 using NVIDIA Grace Blackwell systems. DataCamp and LangChain launched an AI Engineering track.
West Virginia Schools Should Avoid AI and Humanoid Robots
West Virginia's public schools face significant challenges, with many 8th graders unable to perform basic math operations. Despite these issues, introducing AI and humanoid robots into classrooms is not the solution. The humanoid form offers no real advantage, as other robots or devices can perform tasks more efficiently. Furthermore, replacing existing technology like iPads with expensive humanoid robots would be wasteful. The article argues that focusing on personalization through AI might isolate students and misses the broader goals of public education, which include fostering community.
Hardware Root of Trust Crucial for AI Chip Security
The growing demand for AI chips raises concerns about hardware integrity, as counterfeit chips can manipulate data or steal intellectual property. Traditional security methods are insufficient for AI's advanced needs. A hardware root of trust, built into the chip at the factory, is essential for verifying firmware and managing keys, creating an immutable chain of custody. This technology prevents counterfeit chips from running and stops rootkits by ensuring only authentic, signed firmware can be loaded. Implementing a hardware root of trust secures the entire AI lifecycle from manufacturing to deployment.
Take-Two AI Team Disbanded; Head Luke Dicken Departs
Take-Two has reportedly disbanded its AI team, with its head, Luke Dicken, announcing his departure. Dicken, who joined Take-Two in early 2025 after a decade at Zynga, expressed disappointment on LinkedIn, stating his team had developed cutting-edge technology for game development over seven years. Much of the team originated from Zynga's applied AI department, which was integrated after Take-Two acquired the company in 2022. Take-Two declined to comment on the situation, despite CEO Strauss Zelnick's recent confirmation of the company's commitment to exploring generative AI in game development.
Nvidia Invests $2 Billion in Marvell, Boosting NVLink Fusion
Nvidia has invested $2 billion in Marvell Technology, a move that integrates a competitor into Nvidia's NVLink ecosystem. NVLink Fusion is an open rack architecture where custom ASICs can be used, but Nvidia's infrastructure components are required, generating revenue for Nvidia. Marvell, a major supplier of custom AI chips for companies like AWS and Microsoft, will provide custom XPUs and networking for NVLink Fusion. This deal allows Nvidia to absorb the custom ASIC trend rather than fight it. Nvidia is also backing the UALink standard through other partnerships, positioning itself broadly in AI infrastructure.
Musician Will.i.am Finds Teaching AI at ASU Life-Changing
Musician Will.i.am describes his experience teaching a 15-week course on agentic AI at Arizona State University (ASU) as life-changing, even more nerve-wracking than performing at the Super Bowl. The course, titled 'The Agentic Self,' explored ethical AI use and featured guest speakers like the head of ChatGPT. Will.i.am plans to teach another course at ASU and hopes to expand AI education to community colleges. He emphasized the importance of ethical AI use and preparing adults for the rapid advancement of the technology.
Catholic Chatbots Emerge for Evangelization and Faith
The Vatican is navigating the rapid advancement of AI, with Pope Leo XIV emphasizing the need to guide technological innovation responsibly. While AI presents challenges, Roman Catholics, especially laypeople, are increasingly using AI tools for evangelization and catechesis, leading to a rise in 'digital Catholicism.' A Vatican document, 'Antiqua et nova,' stresses the distinction between human and artificial intelligence and the need for 'wisdom of the heart.' Apps like Truthly offer accessible ways to ask questions about Church doctrine, though discussions continue on the appropriate role of AI in spiritual formation.
ZeroEyes AI Now Detects Knives and Tracks People for Security
ZeroEyes has expanded its AI-based threat detection to include knife detection and person tracking, transforming it into a comprehensive physical security intelligence platform. The system integrates with existing security cameras to identify threats like knives and firearms within seconds. Unlike facial recognition, ZeroEyes uses visual characteristics like clothing and movement to track individuals across multiple camera feeds. New analytics features, including drone integration, help reconstruct event timelines. CEO Mike Lahiff aims for ZeroEyes to be the trusted platform for public safety in schools, hospitals, and stadiums.
AI Industry Aims to Automate Its Own Research
Protesters gathered in San Francisco demanding a halt to the development of superintelligent AI, particularly models that can create future AI models. The AI industry is increasingly focused on automating its own research processes, with companies like OpenAI and Google DeepMind developing AI systems that can write code, optimize performance, and conduct literature reviews. While these advancements are significant, humans still play a crucial role in areas like hypothesis generation and resource allocation. Experts believe AI could rapidly improve its own capabilities, potentially leading to faster breakthroughs.
OpenAI's AGI Deployment Head Fidji Simo on Medical Leave
Fidji Simo, OpenAI's head of AGI deployment, is taking a medical leave of absence for several weeks due to a neuroimmune condition. During her absence, OpenAI president Greg Brockman will oversee product development, including super app efforts. CMO Kate Rouch is also stepping down temporarily for health reasons, with Gary Briggs stepping in as interim CMO. COO Brad Lightcap is transitioning to a new role focused on special projects. OpenAI stated its strong leadership team is focused on advancing research, growing its user base, and powering enterprise use cases.
Sources
- RSAC 2026: The AI 'Tragedy of the Commons' and the Future of Agentic Security
- 12 Agentic AI Startups To Watch In 2026
- Artificial Intelligence News for the Week of April 2; Updates from DataCamp, Insight Jam, LearnUpon & More
- AI and humanoids have no place in West Virginia's schools
- Hardware Root of Trust Essential for AI Chip Integrity
- Take-Two Reshuffles Its AI Team: 'It's Truly Disappointing'
- Why Nvidia just poured $2 billion into AI ASIC competitor Marvell — NVLink Fusion turns into soft ecosystem lock-in
- Why will.i.am says teaching at ASU changed his life
- The Rise of the Catholic Chatbot
- ZeroEyes expands AI-based threat detection into the physical realm to secure public spaces
- The AI Industry Wants to Automate Itself
- OpenAI’s AGI boss is taking a leave of absence
Comments
Please log in to post a comment.