TLDR IT 2026-04-24

Google is positioning data as the missing layer for enterprise AI, introducing an Agentic Data Cloud that connects structured and unstructured data to give AI agents real context. This signals that success with agents will depend less on models and more on data architecture, governance, and interoperability across systems.

Researchers identified a malicious npm supply-chain attack targeting Namastex Labs packages. The worm-like malware steals developer credentials, API keys, and cryptocurrency wallets, then self-propagates by injecting malicious code into new package versions. It also targets PyPI, requiring immediate secret rotation and removal of compromised dependencies from CI/CD pipelines.

Australia is working with Anthropic and other software companies after Anthropic's limited-release Mythos cybersecurity model surfaced thousands of significant vulnerabilities across major operating systems and web browsers. The model is intended for defensive security work, but its autonomous coding capabilities are raising concerns that similar tools could also accelerate sophisticated attacks.

AI projects are stalling as security, legal, and compliance teams slow adoption due to concerns around data leakage, model risk, and unclear regulations, creating friction during procurement and deployment. The core issue is that existing governance models weren't built for AI, forcing organizations to rethink how they balance speed, risk, and cross-functional decision-making.

Enterprise AI agents often fail because they lack the necessary context, not because of model limitations. Data products, which package schema, lineage, and semantics as first-class assets, provide the required infrastructure. Adopting this architectural layer ensures agents remain reliable, auditable, and scalable across complex organizational data environments.

Ramp uses Onyx, the open-source AI platform that indexes your company data across Drive, Slack, Notion, and Confluence to provide massively better answers. No MCPs, no search APIs. Just a self-hosted AI coworker that uses your data to provide the info you and your customers need. Work with AI that understands your company

SAP and Google Cloud are integrating Gemini AI into core business processes through a new Unified Data Foundation. This partnership enables zero-copy data sharing via BigQuery, reducing TCO by 54% while mitigating AI hallucinations. These tools empower enterprises to deploy autonomous agents for complex, multi-step task execution.

Microsoft Discovery provides an agentic AI platform for R&D, integrating reasoning, high-performance computing, and Azure security. Partners like Syensqo, PhysicsX, and Synopsys use the platform to automate discovery loops, optimize complex engineering designs, and accelerate innovation cycles, enabling teams to move from hypothesis to outcome with increased confidence.

Band emerged from stealth with $17M to build a communication and orchestration layer that lets AI agents across different frameworks, clouds, and tools discover each other, delegate tasks, and collaborate in real time. The platform introduces an agentic mesh and control plane to manage context, permissions, and routing, aiming to replace brittle integrations with a unified system for multi-agent workflows.

Passkeys provide robust authentication via WebAuthn, eliminating shared secrets and phishing risks. However, they do not secure the application session itself. Developers must still mitigate XSS and CSRF risks using Content Security Policy, strict cookie attributes, and fresh authentication challenges for high-risk operations to ensure comprehensive security.

Nation-state actors are building large-scale covert networks from compromised edge and IoT devices to enable stealthy persistence and potential disruption. The activity focuses on maintaining long-term access across infrastructure, making visibility into edge devices, remote access paths, and unmanaged assets a growing blind spot for enterprise IT.

OpenAI Privacy Filter is an open-weight model for detecting and redacting personally identifiable information (PII) in text with state-of-the-art accuracy.

Arista shares hit a 52-week high after Google detailed a new AI data center networking architecture.