Here's today's top stories in AI news. 

OpenAI launched Shopping Research in ChatGPT, a tool that builds tailored, research-grade buying guides with product comparisons and current retailer info; it appears on relevant prompts, takes a few minutes to generate, and will later tie into Instant Checkout. 

Google’s AI reset continues with Gemini 3 debuting directly inside Search; while the company touts record results and end-to-end stack advantages, AI Overviews and conversational search raise unresolved questions about ad economics and the web’s health, with publishers reporting traffic strain and Google disputing studies showing reduced clicks. 

New research from Cardiff University and Ca’ Foscari finds LLMs recognize pun structures without true understanding; performance drops sharply on unfamiliar wordplay, underscoring limits in humor and cultural nuance. 

Sunday Robotics emerged with Memo, a two-armed home robot trained on 10 million real routines via $200 “Skill Capture Gloves,” handling dishes, laundry, and coffee with zero-shot generalization; pilot shipments target 2026. 

Anthropic released Opus 4.5, setting new coding and tool-use benchmarks, adding improved long-context memory and “endless chat,” and expanding Claude for Chrome and Excel.

**Summary:** OpenAI is acquiring Jony Ive's design startup, Io, for $6.5 billion to produce hardware specifically for ChatGPT, aiming to create devices that reduce friction between users and AI, potentially rivaling major tech firms. Concurrently, Check Point reports a Facebook ad campaign impersonating Kling AI, exposing 22 million users to malware. In finance, AI automation is driving lower losses and faster growth in lending platforms, with firms like Upstart reporting significant increases in loan originations. Meanwhile, a grading incident at a Maine high school highlights regulatory gaps regarding AI use by teachers, contrasting strict student bans. Finally, Klarna showcased a lifelike AI avatar of its CEO in its earnings video as part of its IPO preparations, reflecting discussions on AI's role in leadership roles.

OpenAI has launched its GPT-4.1 model on ChatGPT, accessible to both free and paid users, enhancing speed and accuracy while offering free users usage limits and paid users priority access. This move aims to democratize advanced AI tools amidst growing market competition. Tencent leverages its WeChat ecosystem of 1.4 billion users to enhance AI adoption by integrating features into messaging, payments, and video, thereby capitalizing on user data and solidifying its dominance in China's digital economy. YouTube is rolling out an AI feature using Google's Gemini to optimize ad placements based on real-time viewer engagement, aiming to improve ad effectiveness and audience targeting without increasing ad volume. A study from Queensland University of Technology warns that AI job interviews can perpetuate discrimination and biases in recruitment, challenging the notion that automation leads to fair hiring practices. An article in AI and Ethics argues that current AI systems may exhibit a form of free will, urging a reevaluation of AI autonomy and its implications for regulation and accountability.

Nvidia and AMD have secured key contracts to supply AI chips to Saudi Arabia's Humain for a major AI data center project, backed by the Public Investment Fund. This initiative will involve delivering hundreds of thousands of GPUs over five years and aims to build a 1.9-gigawatt data center by 2030. This deal signals a change in U.S. AI chip export policy, boosting Nvidia's stock. OpenAI launched HealthBench, a benchmark for evaluating AI in healthcare, developed with input from 262 physicians. It aims to validate AI models against real-world medical standards. Audible is expanding its audiobook offerings through AI production technology, enabling publishers to create audiobooks in multiple languages with AI-generated voices, enhancing global access. Andreessen Horowitz partners discussed the ambiguity around 'AI agents,' noting current limitations in autonomy and memory, indicating a need for clearer definitions. Stella Garber proposed a five-question framework for assessing AI startup ideas, emphasizing the importance of practical execution and aligning solutions with real-world needs.

**Perplexity AI's Valuation Soars to $14 Billion Amid New Funding:** Perplexity AI has secured a $500 million funding round led by Accel, boosting its valuation from $9 billion to $14 billion. The San Francisco startup, which offers an AI-driven search tool, plans to launch its own web browser, Comet, intensifying competition with traditional search engines. Despite strong investor interest, including from Nvidia and Jeff Bezos, it faces legal issues over content usage. **Stargate's $100B AI Ambition Hits Tariff Roadblocks:** OpenAI and SoftBank's $100 billion Stargate project is stalled due to funding delays and new U.S. tariffs raising data center construction costs. While Oracle's development in Texas progresses, the overall project faces uncertainty amid economic challenges. **Hackers Spread Malware by Mimicking Popular AI Tools:** Cybercriminals are using fake AI tools to distribute malware that harvests sensitive data. Victims are tricked into downloading the malware disguised as legitimate software, highlighting the need for vigilance against sophisticated online threats. **AI Assistants Are Quietly Reshaping the Doctor-Patient Experience:** Healthcare providers are increasingly using AI phone assistants to automate appointment scheduling, managing over 70% of calls. This improves efficiency and patient care while allowing staff to focus on complex tasks. **Zuckerberg’s AI Friends Push Confuses Technology With Human Relationships:** Mark Zuckerberg suggests AI chatbots could help combat social isolation, but critics argue they can't replace genuine human relationships. The debate raises ethical concerns about substituting human companionship with AI, stressing the importance of authentic social connections.