Google AI Updates: May 22, 2026

1. Google Beam adds life-size group meetings on HP Dimension with spatial audio

Google. Google Research detailed a new experiment that extends Google Beam, its volumetric video platform, to group meetings on the HP Dimension display, rendering remote participants at true-to-life size and seating them as if they were at a shared table. The system anchors each speaker’s voice in space using spatial audio and auto-tunes itself for both home and office rooms, and Google cited internal research showing a 50% stronger sense of social connection and a 21% increase in participants’ ability to contribute to the conversation versus standard video calls. Beam works alongside Google Workspace and Zoom, and the group-meetings experiment is rolling out as the next step beyond the one-to-one Beam calls that shipped earlier this year. Source

2. Google DeepMind opens an APAC accelerator for AI-driven environmental work

Google. Google DeepMind launched an Asia Pacific edition of its accelerator program, framed around “AI for the Planet” and aimed at startups, research teams, and nonprofits working on nature, climate, agriculture, and energy. The three-month program kicks off with an in-person bootcamp in Singapore and pairs cohort teams with DeepMind mentors plus access to Google’s frontier AI and science AI models, building on the company’s existing AlphaEarth, AlphaFold, and Co-Scientist work. The launch is timed against a KPMG-Google report finding that green technologies in APAC are not scaling fast enough to keep up with the region’s environmental vulnerabilities, and positions DeepMind’s accelerator network as a counterpart to Google.org’s existing AI for Social Good funding rather than a replacement. Source

3. Google Lighthouse adds an Agentic Browsing audit that checks for llms.txt and machine-readable metadata

Google. Google rolled out an experimental Agentic Browsing category in Lighthouse that scores how well websites expose structured signals to AI agents, including detection of the emerging llms.txt convention, schema.org markup completeness, and machine-readable robots directives. The audit surfaces per-page scores in DevTools alongside existing performance, accessibility, and SEO categories, and gives publishers a concrete, Google-blessed checklist for agent-grade indexing rather than the ad-hoc guidance that had circulated through community blog posts. The move effectively nudges llms.txt toward de facto standard status and signals that Google is preparing search to be queried by agents at least as often as by humans. Source