Good morning. It's Tuesday, September 22, and we're covering Google’s Gemini-powered laptop push, the UN’s call for stronger AI safeguards, and Apple’s costly settlement over its delayed Siri AI features.

Plus: a guest essay on how AI agents are transforming travel in Tokyo.

Apple’s delayed Siri AI rollout could now cost the company millions in consumer payouts. A $250 million settlement lets eligible U.S. buyers claim compensation for iPhone 15 Pro and iPhone 16 models purchased from June 10, 2024, to March 29, 2025. Payments start at $25 and can reach $95 per device. Claims close Dec. 21, 2026.

Spanish Prime Minister Pedro Sanchez says AI companies cannot regulate themselves as governments address privacy, discrimination and misinformation risks. Spain’s 12-month roadmap includes an AI Gigafactory and models developed with the Barcelona Supercomputing Centre for climate, health and energy. Sanchez also calls for stricter data-center energy and environmental standards.

Grok 4.7 targets longer coding and knowledge-work tasks with a larger base model and extended reinforcement learning. xAI says it matches Grok 4.6’s speed while pricing API access at $2 per million input tokens and $6 per million output tokens. The model scores 46.3% on CursorBench 4.0 and 71% on DeepSWE v1.1.

Enterprise AI is only as valuable as the context behind it. Now, you can securely engage with Box content within ChatGPT, bringing enterprise content directly into AI workspaces. The embedded Box experience makes it easy to find, preview, and work on enterprise content, without searching for files or moving content out of Box.

Bring your enterprise’s knowledge into AI-powered workflows.

_{^{Guest contributor: Alex Salkever is the Editor-in-Chief of the Agentic AI Foundation and the Linux Foundation, where he is responsible for storytelling across agentic AI and all of open source.}}

I'm writing this to you from the Agentic future on the train to Narita Airport from Tokyo after an eye-opening trip. I have receipts. Perhaps the best way to understand the power of agentic AI (and AI in general) is to travel to a country where you don't speak the language, and most people there don't speak it either. → Read the full article here.

Google opened preorders September 21, 2026, for its $899 Googlebook, an Android-based laptop designed around Gemini AI and desktop computing. The devices add features including Magic Cursor, which connects selected screen content to Gemini, and Rambler, which turns rough voice dictation into polished text, alongside AI-powered widgets.

Google is positioning the Googlebook as a step beyond traditional Chromebooks, with the longer-term possibility of moving its large education customer base toward Gemini; the company says roughly 50 million Chromebooks are used in schools. The laptops include 12 months of Google AI Pro and will begin shipping October 4 in the U.S. → Read the full article here.

A United Nations scientific panel is urging governments to strengthen safeguards for increasingly capable AI agents before researchers can fully establish how emerging risks occur. Its first thematic brief argues that potential loss-of-control scenarios could cause irreversible harm, making scientific uncertainty insufficient grounds for delaying protections.

The report arrives as world leaders gather in New York for the UN General Assembly and the U.S. and China prepare for AI discussions. The panel also calls for greater international coordination on AI safety and accountability, noting that countries may pursue different regulatory approaches. → Read the full article here.

Amazon has blocked Meta’s Muse AI assistant from accessing Amazon.com for shopping, telling users that continued access by an “unauthorized AI agent” violates its Conditions of Use. The move prevents customers from completing Amazon purchases through Muse and highlights an emerging friction point as AI agents begin acting on users’ behalf online.

Amazon operates its own AI models and infrastructure, but agent-driven shopping also creates practical risks: if an AI makes a bad purchase, Amazon may face the resulting customer and seller disputes. Muse reportedly has a relatively low hallucination rate compared with other AI systems, but it remains imperfect, giving Amazon another reason to limit automated purchasing for now. → Read the full article here.

OpenAI Proposes AI Safety Standards: OpenAI released international safety standards as U.S. and China weigh cooperation on AI risks.

Samsung Backs Kairos Reactor: Samsung C&T will invest up to $100M and provide engineering services for Kairos’ 50-megawatt Google reactor.

Tabby Targets Accounting Work: Former accountant Ahad Ali built Tabby to automate bookkeeping and deliver real-time profit-and-loss data.

ByteDance Launches Dramagic: The AI platform handles scripts, characters, storyboards and video previews, with collaboration and consistency checks.

SoftBank Launches $11B Bond Sale: SoftBank is raising $10B in dollar bonds and €1B in euro notes to fund its OpenAI investment.

Thanks for reading. See you next time!

— Matthew Berman, Nick Wentz & the Forward Future Team