AI News Roundup — April 17, 2026
Google rolls out a major AI Mode update for Chrome featuring side-by-side browsing and deep Gemini integration, while a new investigation reveals how federal agencies are using AI to supercharge surveillance of Americans. NVIDIA releases Lyra 2.0 for persistent AI-generated 3D worlds, and a YC-backed startup launches the first marketplace for AI agents to route tasks to human experts.
Google Deeply Integrates AI Mode into Chrome
Google rolled out a major AI Mode update for Chrome on desktop and mobile. The headline feature is side-by-side browsing — clicking a link from AI Mode opens the page alongside the AI panel so users can explore and ask follow-up questions without losing context. The Chrome new tab page also receives an inline AI prompt box with the ability to attach recent tabs, upload files and images, generate images, and switch between Gemini 3 models. The update is live in the US with international rollout planned.
Report: AI Is Supercharging Federal Surveillance of Americans
A new US News investigation examines how federal agencies are leveraging AI to dramatically expand surveillance capabilities — from facial recognition and predictive policing to social media monitoring at a scale previously impossible. The report raises significant alarm about the absence of legal guardrails governing AI-driven government surveillance of US citizens.
NVIDIA Releases Lyra 2.0 for Persistent AI-Generated 3D Worlds
NVIDIA launched Lyra 2.0, a major update to its AI world-generation platform that resolves the longstanding drift and consistency problems in AI-generated 3D environments. The result is persistent, explorable 3D worlds that remain coherent across sessions — a significant advance for gaming, simulation, virtual reality, and synthetic training data generation.
Google Releases Gemini 3.1 Flash TTS with 70-Language Support
Google launched Gemini 3.1 Flash TTS, a new text-to-speech model with scene direction capabilities, speaker-level control, and support for 70 languages. The model enables developers to craft nuanced multi-speaker audio with fine-grained control over tone and delivery, raising the bar for AI-generated voice and audio applications.
TeraWulf Raises $900 Million to Build Kentucky AI Data Center Campus
TeraWulf priced an upsized $900 million equity offering at $19 per share to fund its Hawesville, Kentucky AI data center campus. The raise underscores the continued massive capital flows into AI infrastructure as demand for GPU compute continues to dramatically outpace available supply across the industry.
YC-Backed Humwork Launches First AI-to-Human Expert Marketplace
Humwork, backed by Y Combinator, launched what it describes as the first A2P (Agent-to-Person) marketplace — a platform where AI agents can route tasks to verified human domain experts via MCP in under 30 seconds. The model inverts the conventional human-assigns-AI paradigm, letting autonomous agents decide when and how to escalate to human judgment and expertise.
Chrome Zero-Day CVE-2026-5281 Under Active Exploitation — Patch Released
Google disclosed and patched a critical Chrome zero-day (CVE-2026-5281) that is already being actively exploited in the wild. All Chrome users on desktop and mobile are urged to update immediately. The vulnerability is especially notable given the browser's newly expanded AI Mode integration, which significantly broadens its attack surface.