google launches amazon while nvidia expands its platform

Key Takeaways

• AI agents pose significant security risks, including prompt injection and scattered credentials, with 51% of external actions using hard-coded passwords.
• An identity gap hinders the safe scaling of AI agents, necessitating distinct identity management for effective policy enforcement and incident investigation.
• Aembit launched an Identity and Access Management (IAM) platform for agentic AI, featuring 'Blended Identity' to verify both user and agent for granular access control.
• Google's Gemini app now includes 'notebooks' for Ultra, Pro, and Plus subscribers to organize chats and files, syncing across Gemini and NotebookLM.
• Twinnin, backed by Google and Nvidia, is a platform cloning actors' faces for ethical monetization, but faces industry debate over consent and potential misuse.
• AWS and TM Forum are collaborating to scale trusted AI for AI-native 6G networks, focusing on unified data foundations and experiential validation.
• Nvidia's own applied deep learning research teams are experiencing a GPU shortage, prompting a focus on GPU-efficient models like Nemotron to manage demand.
• AI adoption in CRM is growing, with 45% of sales professionals using AI weekly, leading to shorter deal cycles and increased win rates, though data quality is critical.
• Avalara is embedding AI agents for AI-executed compliance workflows, automating tasks like tax calculations and exemption certificate management.
• 80% of white-collar workers resist AI adoption due to fear and a lack of relevant training, highlighting a trust gap between employees and executives.

AI agents pose security risks, CEO warns

Itamar Apelblat, CEO of Token Security, highlights growing security dangers from AI agents, such as scattered credentials and prompt injection. He notes that 51% of external AI agent actions still use hard-coded passwords, and a single bad prompt can bypass security. Apelblat stresses the need to manage AI agent intent like policy, ensuring security measures adapt even when users change their requests. Many organizations treat AI agents carelessly, similar to early cloud service accounts, leading to unclear ownership and lingering access.

Identity gap hinders AI agent scaling

A significant identity gap is preventing AI agents from being safely scaled in production environments. Current systems struggle to identify AI agents as distinct from humans or traditional machines, as they operate continuously and make unpredictable decisions. Experts emphasize that identity must be a primary requirement, ensuring every actor, including AI agents, has a clear identity to enable effective policy enforcement and incident investigation. Without this, autonomous AI agents with broad privileges become high-value targets for attackers, leading to compromised systems and data leaks.

Aembit launches AI agent security platform

Aembit has launched its Identity and Access Management IAM platform for agentic AI, addressing the gap in securing autonomous systems. The platform includes an Authorization Server and Identity Gateway to manage non-human identities and workload access. It introduces 'Blended Identity,' a new model that verifies both the user and the agent for resource access, ensuring granular control and per-user credential isolation. This system aims to provide production-grade security infrastructure for AI agents already operating in enterprise environments.

AI actor platform Twinnin sparks debate

The new platform Twinnin, which clones actors' faces to protect and monetize their identity in the age of AI, has divided the UK film and TV industry. Backed by Google and Nvidia, Twinnin uses 'immutable provenance technology' to create digital likenesses that can be licensed to studios and brands. While founder Katrien Grobler believes it empowers actors to own their identity, unions like Equity and some actors express concerns about consent, potential misuse, and safeguarding, especially for child actors. The platform aims to provide ethical and consensual licensing of digital likenesses.

Gemini app adds notebooks for project organization

Google's Gemini app now features 'notebooks' to help users organize chats and files for complex projects. These notebooks act as personal knowledge bases that sync across the Gemini app and NotebookLM, offering enhanced learning and working features. Available this week for Google AI Ultra, Pro, and Plus subscribers on the web, notebooks allow users to consolidate conversations, provide custom instructions, and upload files for better AI context. This feature aims to streamline workflows and improve the organization of information for users.

AI in CRM: Adoption grows but data quality is key

AI adoption in CRM is rapidly increasing, with 45% of sales professionals now using AI weekly, leading to shorter deal cycles and increased win rates. However, the effectiveness of AI in CRM heavily relies on the quality of the underlying data. Experts warn that incomplete or messy CRM data limits AI's potential, emphasizing disciplined data practices over flashy add-ons. While AI boosts productivity, some sales teams fear job displacement, highlighting a mixed reception despite clear benefits like increased revenue and efficiency.

AWS and TM Forum collaborate on AI-native 6G networks

AWS and the TM Forum are collaborating to address challenges in scaling trusted AI and data products for AI-native 6G networks. Key barriers include data fragmentation across proprietary systems, the 'Everywhere Problem' of ensuring reliable performance across diverse environments, and the need for sustainable scaling for economic viability. The initiative focuses on creating unified data foundations, enabling experiential validation through simulations and digital twins, and fostering collaboration to build secure, scalable AI systems for future networks.

Workers resist AI adoption due to fear

A new global survey reveals that 80% of white-collar workers are resisting AI adoption mandates, not due to lack of functionality, but fear of its potential. A significant trust gap exists between employees and executives regarding AI's capabilities and impact on roles. Many workers feel unsupported and that AI training is irrelevant to their jobs, leading to lost productivity. The report highlights that while AI can save time for proficient users, it destroys productivity for those who struggle with it, indicating a need for better training and clear policies.

Avalara shifts to AI-executed compliance workflows

Avalara is transforming compliance by embedding AI agents directly into its platform, moving from AI-assisted tasks to AI-executed workflows. This includes automating tax calculations, return preparation, and exemption certificate management. Leveraging its ALFA framework and MCP servers, Avalara's AI agents operate continuously and accurately within a human-approved, audit-ready environment. This approach aims to reduce risk, improve accuracy, and provide greater confidence in transactions by making compliance an AI-executed system.

AI GPU shortage impacts Nvidia research teams

Even Nvidia's own applied deep learning research teams are struggling to access enough GPUs due to the high demand for AI computing power. Bryan Catanzaro, VP of applied deep learning research at Nvidia, stated that securing more GPUs is a primary job for him as teams are supply constrained. This scarcity is driving Nvidia to focus on GPU-efficient models like Nemotron to maximize the use of available resources. The company is taking a more active role in shaping the AI ecosystem to ensure its hardware and software remain central.

AI's impact on smart homes discussed at conference

The 30th annual CONNECTIONS Conference, held May 5-7 in Santa Clara, California, will bring together connected home executives to explore the future of AI, smart home technology, and security. Speakers from companies like Alarm.com, Silicon Labs, AWS, and Ring will discuss innovations in AI, intelligent consumer services, and platforms. The conference aims to highlight the growing impact of artificial intelligence on the connected home and related industries.

AI training lags behind employee priorities

A Docebo report indicates that while AI literacy and skills are top priorities for many employees, their organizations' training efforts are falling short. Many employees feel unsupported and that current AI training is not relevant to their specific roles or career progression. Despite high demand for AI readiness, companies struggle to provide tailored training that connects AI skills to personal development. Experts suggest that effective training must be role-specific and include human skills like critical thinking to improve AI ROI and employee confidence.

Sources

• AI agent intent is a starting point, not a security strategy
• The Identity Gap Blocking Agentic AI at Scale
• Aembit IAM for Agentic AI Is Now Generally Available
• AI Actors Platform Twinnin Divides Opinion In Film & TV Industry
• Even Nvidia’s own research teams can’t get enough GPUs amid the race for AI computing power
• Try notebooks in Gemini to easily keep track of projects
• The State of AI in CRM: A Sneak Peek Into 2026
• Scaling trusted AI and data products: shaping new pathways for AI-native 6G
• White-collar workers are quietly rebelling against AI as 80% outright refuse adoption mandates
• Avalara Advances Agentic Compliance, Moving from AI-Assisted to AI-Executed Workflows
• Parks Associates CONNECTIONS Conference Highlights Impact of AI
• AI literacy tops learning priorities but training efforts lag