Latest AI News

AI-linked layoffs aren't reliably boosting stock prices, according to a CNBC analysis of 23 S&P 500 companies. As of May 15, 56% of those firms traded in the red after announcing AI-related workforce cuts, with an average decline of about 25%. Nike fell nearly 35% after cutting 775 workers to accelerate automation, Salesforce dropped 32% after eliminating 4,000 jobs citing its Agentforce AI bots, and Fiverr plunged 54% after laying off 30% of staff to become an AI-first company.

Former Google CEO Eric Schmidt was booed by students at the University of Arizona during a commencement speech after saying AI will touch every profession, classroom, hospital, and laboratory. Schmidt acknowledged the crowd's reaction, noting fears that jobs are evaporating and the future has been written. The incident is at least the third commencement speech this month to draw jeers over AI remarks, following similar responses at the University of Central Florida and Middle Tennessee State University.

OpenAI and the Government of Malta have announced a world-first partnership to provide ChatGPT Plus to all Maltese citizens. Under Malta's AI for All initiative, citizens complete a free AI literacy course developed by the University of Malta, then receive one year of ChatGPT Plus at no cost. The Malta Digital Innovation Authority will manage distribution. The program launches in May 2026 and is part of OpenAI's broader OpenAI for Countries initiative, which also includes partnerships with Estonia and Greece.

Anthropic has published a deployment guide for Claude across the legal industry, as generative AI adoption among general counsel doubled to 87% in 2026, up from 44% the prior year, according to an FTI Consulting/Relativity report. The guide covers four products — Claude Chat, Claude Cowork, Claude for Microsoft 365, and Claude Platform — plus 12 pre-built practice-area plugins spanning Commercial, Corporate, Privacy, Litigation, and AI Governance work, alongside a three-phase adoption roadmap for firms and in-house teams.

Anthropic and PwC have expanded their strategic alliance, deploying Claude across agentic software development, AI-native deal-making, and enterprise function reinvention. PwC will roll out Claude Code and Claude Cowork to hundreds of thousands of professionals globally, backed by a joint Center of Excellence and a program to train and certify 30,000 U.S. staff. PwC is also launching an Office of the CFO business unit built on Claude. Live deployments have cut delivery times by up to 70%, compressing insurance underwriting from ten weeks to ten days.

OpenAI has launched a personal finance preview in ChatGPT for U.S. Pro subscribers, allowing users to securely connect accounts from over 12,000 financial institutions via Plaid, with Intuit support coming soon. The feature provides a spending dashboard and lets users ask questions grounded in their real financial data, powered by GPT-5.5's reasoning capabilities. Financial memories store user goals across conversations. OpenAI plans to expand access to Plus users and eventually everyone after gathering early feedback.

xAI now lets Grok subscribers use their accounts inside Nous Research's open-source Hermes Agent, a self-improving agent that runs persistently on any computer, sandbox, or VPS and builds long-term memory across sessions. The integration gives Hermes access to Grok 4.3 for text and advanced reasoning, Grok Text-to-Speech for spoken responses, and Grok Imagine for image and video creation. Hermes also connects to WhatsApp, Discord, Telegram, and Signal. The feature is available on every Grok subscription tier.

California resident Amargo Couture has filed a class-action complaint against OpenAI Global LLC in the Southern District of California, alleging the company embedded Meta's Facebook Pixel and Google Analytics into the ChatGPT web interface. The suit claims these integrations transmitted user chatbot queries and associated data into advertising ecosystems. Bloomberg Law also reported on the case. OpenAI has not issued a public statement. The lawsuit highlights growing legal scrutiny over telemetry practices in AI chat products.

Meta has opened a developer preview for building apps on Meta Ray-Ban Display glasses, offering two build paths: the Meta Wearables Device Access Toolkit, a native iOS and Android SDK supporting text, images, and video on the display, and a Web Apps path using standard HTML, CSS, and JavaScript. Glasses also support gesture controls via the Meta Neural Band using surface electromyography. Developers can share builds with up to 100 testers during the preview period.

Perplexity has launched a Snowflake App Connector for its Computer product, letting teams query enterprise data in plain language without writing SQL. Available to Pro, Max, and Enterprise users, the connector allows questions like forecasting demand or identifying top-performing marketing campaigns. Admins retain control via role-based access, read-only permissions, and User OAuth. The idea came from Perplexity's internal Slackbot, which now handles up to 3,000 weekly queries. Data is never used to train models.

A developer named Neilson has open-sourced a 3D generation toolkit built specifically for use with Anthropic's Claude Code. The tool accepts a single input image and automatically generates complete 3D environments, producing meshes, physics simulations, lighting setups, and audio components from that one source. By releasing the toolkit publicly, Neilson aims to make it significantly easier for developers working within Claude Code to rapidly build and deploy fully realized 3D scenes without manual asset creation.

Researchers at Palo Alto-based security firm Calif used an early version of Anthropic's unreleased Mythos AI model, also called Claude Mythos Preview, in April to uncover significant bugs in Apple's macOS. The team chained two vulnerabilities with advanced techniques to corrupt Mac memory and access restricted system areas. The work is part of Anthropic's Project Glasswing, which shares Mythos with select partners including Apple, Google, and Microsoft to proactively find and patch zero-day vulnerabilities before malicious actors can exploit them.

xAI has launched an early beta of Grok Build, an agentic command-line interface tool built for coding, building applications, and automating workflows. The tool is currently available exclusively to SuperGrok Heavy subscribers, xAI's premium tier. During this beta period, xAI plans to refine both the underlying model and the product itself based on user feedback. Developers interested in trying Grok Build can access it now at grok.com.

Anthropic and the Gates Foundation have announced a $200 million partnership combining grant funding, Claude usage credits, and technical support across global health, education, and economic mobility over four years. Key initiatives include accelerating vaccine and therapy development for neglected diseases like polio, HPV, and preeclampsia, improving disease forecasting with the Institute for Disease Modeling, building AI tutoring tools for K-12 students in the US, sub-Saharan Africa, and India, and supporting smallholder farmers through agriculture-specific Claude improvements.

A live-action AI film directed by Dustin Yellin, starring Paul Rudd and Chris Rock, and produced by Darren Aronofsky will premiere at the Cannes Film Festival next week. Nick St. Pierre, who shared the announcement on X, is an executive producer on the project. The film blends live-action footage with AI-generated elements and marks a notable moment for AI in mainstream cinema, with high-profile talent attached to the production.

Google is set to unveil Gemini Spark at Google I/O, a new 24/7 AI agent designed to function as an everyday personal assistant. Gemini Spark will learn from user behavior and integrate with connected apps and skills, allowing it to handle tasks autonomously on an ongoing basis. The announcement positions Gemini Spark as a persistent, proactive agent rather than a reactive chatbot, expanding Google's AI assistant capabilities significantly.

Google Gemini 3.2 Flash is rumored to achieve 92% of GPT-5.5's performance on coding and reasoning benchmarks while costing 15 to 20 times less on inference. According to Bindu Reddy on X, the model leverages DeepMind's distillation techniques to deliver these gains. Latency improvements are also notable, with response times reportedly dropping below 200 milliseconds, making it a compelling option for cost-sensitive, performance-demanding applications.

Figure, the humanoid robotics company founded by Brett Adcock, completed and live-streamed eight consecutive hours of fully autonomous, unsupervised robot operation with no human intervention required. Following the milestone, Adcock announced the company is moving beyond the single session to launch a 24/7 continuous livestream, allowing the public to watch Figure's robots perform autonomous work around the clock without stopping.

NV Energy will cut 75 percent of Liberty Utilities' electricity supply to the California side of Lake Tahoe by May 2027, affecting about 49,000 customers, to redirect power to AI data centers in Nevada. Companies including Google, Apple, and Microsoft are building large facilities near Reno. Residents report electricity prices have risen 77 percent since late 2022. Liberty Utilities says it filed with the CPUC in March 2026 to find new wholesale power suppliers, with a formal request for proposals expected this summer.

Perplexity has detailed the security architecture behind Perplexity Computer, its autonomous agent that writes code, browses the web, and connects to external services. Each task runs inside a Firecracker microVM sandbox with a dedicated Linux kernel, isolated filesystem, and private network namespace. The system includes a four-layer prompt injection defense using the BrowseSafe detection model, audited by Trail of Bits. Enterprise features include SAML SSO, audit logs integrating with Splunk and Datadog, and granular connector controls for services like Gmail, Slack, and Salesforce.

OpenAI engineer David Wiesen built a custom Windows sandbox for Codex, the company's coding agent, after existing Windows tools like AppContainer, Windows Sandbox, and Mandatory Integrity Control labeling all fell short. Without a sandbox, Windows users had to either approve nearly every command or enable unrestricted Full Access mode. The solution uses synthetic SIDs and write-restricted tokens to control file writes, while poisoning network escape hatches via proxy environment variables, bringing Windows users the same safety guarantees already available on macOS and Linux.

Anthropic has temporarily increased Claude Code weekly usage limits by 50%, effective immediately and running through July 13. The boost applies to all Pro, Max, Team, and seat-based Enterprise subscribers. Claude Code is Anthropic's AI-powered coding assistant, and the expanded limits give developers significantly more capacity during the promotional window. The increase is already live, meaning users can take advantage of the higher limits right away without any additional steps or configuration changes required.

TikTok has launched the TikTok Ads MCP Server, a model context protocol tool that lets marketers connect AI agents directly to its ads platform to plan, launch, and optimize campaigns without manual intervention. Announced at TikTok World by global head of product marketing Jose Villalobos, the server automates tasks like setting creatives, adjusting bids, and shifting budgets. Google, Meta, and Amazon have launched similar MCP servers, reflecting a broader industry shift toward AI-run advertising operations.

Amazon has merged its Rufus shopping AI and Alexa+ into a single assistant called Alexa for Shopping, now available to all U.S. customers on the Amazon Shopping app, website, and Echo Show devices — no Prime membership required. The assistant combines Rufus's product expertise with Alexa+'s personalized context, enabling features like AI-powered search, dynamic product comparisons, up to one year of price history, scheduled purchases, and agentic buying from third-party retailers via the existing Buy for Me feature.

Meta has launched Incognito Chat with Meta AI on WhatsApp and the Meta AI app, offering truly private AI conversations where not even Meta can read the exchanges. Built on WhatsApp's Private Processing technology, messages are handled in a secure environment, are not saved, and disappear by default. Unlike other incognito-style modes that still log queries, this feature is designed for sensitive topics like health, finances, or career advice. A related Side Chat feature is also planned.

Microsoft has updated Edge for desktop and mobile with new Copilot features, including multi-tab reasoning, browsing history context, long-term memory, and hands-free Voice and Vision. The Edge mobile app gains Journeys, which organizes browsing history into topic cards with summaries and suggested next steps. New desktop tools include Study and Learn mode, a Writing assistant, Copilot quizzes, and a podcast generator. A redesigned new tab page is also rolling out. Microsoft is retiring Copilot Mode as part of the update.

Notion has launched its new Developer Platform, introducing a comprehensive suite of tools designed for developers and AI agent builders. The platform includes a CLI tool called ntn that brings Notion directly into the terminal, Workers for running code on Notion's own infrastructure, database sync functionality to pull in data from any external source, agent tools for building custom workflows, and webhook triggers that allow external applications to connect with and trigger actions inside Notion.

Anthropic launched Claude for Small Business, a package of app connectors and 15 ready-to-run agentic workflows built into Claude Cowork. It integrates with QuickBooks, PayPal, HubSpot, Canva, DocuSign, Google Workspace, and Microsoft 365 to handle payroll planning, monthly close, invoice chasing, and campaign creation. Anthropic also partnered with PayPal on a free AI Fluency course and launched a multi-city SMB Tour starting May 14 in Chicago, offering hands-on training and one-month Claude Max subscriptions.

Recursive has raised $650 million in funding at a $4.65 billion valuation, positioning the company to develop self-improving artificial intelligence systems. The substantial funding round signals strong investor confidence in recursive self-improvement as a viable AI research direction. The company, operating under the handle Recursive SI, aims to build AI capable of autonomously enhancing its own capabilities, a goal that places it among the most ambitiously funded AI startups currently operating in the sector.

Google DeepMind is reimagining the mouse pointer using Gemini AI, enabling users to point at on-screen content and issue natural voice or text commands like "Fix this" or "Move that" without detailed prompts. The experimental system captures visual and semantic context around the cursor, turning pixels into actionable entities. Google is already integrating these principles into Chrome via Gemini in Chrome and plans to launch Magic Pointer on its new Googlebook laptop. Demos are available in Google AI Studio.

Meta has launched Muse Spark, the first model from its Meta Superintelligence Labs, now rolling out across the Meta AI app, WhatsApp, Instagram, Facebook, Messenger, Threads, and Ray-Ban Meta and Oakley Meta smart glasses in the US and Canada. Built from the ground up over nine months, Muse Spark supports complex reasoning, multimodal perception, parallel subagents, and a new shopping mode that searches Facebook Marketplace alongside broader web listings. A private API preview is available to select partners.

OpenAI disclosed that two employee devices were compromised on May 11, 2026, as part of the Mini Shai-Hulud supply chain attack targeting the TanStack npm library. Limited credential material was exfiltrated from internal source code repositories, including code-signing certificates for iOS, macOS, and Windows apps. No user data or production systems were affected. macOS users must update ChatGPT Desktop, Codex App, Codex CLI, and Atlas by June 12, 2026, or their apps will stop functioning when the old certificate is revoked.

Anthropic has surpassed OpenAI in business adoption for the first time, according to Ramp's AI Index for April 2026. Anthropic reached 34.4% of U.S. businesses, up 3.8%, while OpenAI fell 2.9% to 32.3%. Overall AI adoption hit 50.6%. Anthropic quadrupled its business adoption over the past year while OpenAI grew just 0.3%. Headwinds for Anthropic include rising costs, recent outages, and a model update that triples token costs for image-based prompts.

Poppy, a new AI assistant app from San Francisco-based Second Nature Computing, connects your calendar, email, messages, location, and services like Gmail, WhatsApp, and Instacart to proactively surface reminders, suggestions, and tasks. Built by former Humane engineer Sai Kambampati, Poppy can suggest a park walk during a calendar gap or recommend restaurants based on a friend's food preferences. The four-person team raised $1.25 million in pre-seed funding led by Kindred Ventures, with DeepMind's Logan Kilpatrick among angel investors.

RSL Media, a nonprofit cofounded by Cate Blanchett, has launched the Human Consent Standard, an AI licensing framework backed by George Clooney, Tom Hanks, Meryl Streep, Viola Davis, and Kristen Stewart. The standard lets people set terms for how AI systems use their likenesses, creative works, characters, and designs, with options to grant full access, allow access with conditions, or restrict it entirely. A verification registry launches in June, discoverable via robots.txt.

AI developers running agents on MacBooks are popularizing dummy display plugs, a trick suggested by developer Will DePue on X. Plugging a USB-C or HDMI dummy adapter into a MacBook fools it into thinking an external display is connected, enabling clamshell mode and keeping the machine awake without software workarounds. A four-pack of UGreen HDMI dummy adapters costs about nine dollars on Amazon. Alternatives include macOS Terminal's caffeinate command or the free app Amphetamine, but the dongle offers a faster physical toggle.

Google announced Gemini Intelligence for Android at the Android Show 2026, transforming Android from an operating system into an intelligence system. Key features include multi-step task automation on Galaxy S26 and Pixel 10, Gemini in Chrome for browsing and auto-filling forms, Rambler for converting natural speech into polished multilingual text, and Create My Widget for building custom widgets using natural language. Rollout begins this summer on select Samsung Galaxy and Google Pixel phones, expanding to watches, cars, glasses, and laptops later in 2026.

Google has unveiled Googlebook, a new laptop category built around Gemini AI, merging Android and ChromeOS into a single platform. Key features include Magic Pointer, developed with Google DeepMind, which brings contextual Gemini suggestions directly to the cursor, and Create Your Widget, which lets users build personalized dashboards via prompts. Googlebooks also integrate with Android phones for seamless app and file access. Hardware partners include Acer, ASUS, Dell, HP, and Lenovo, with devices launching this fall.

Anthropic has released 20+ new MCP connectors and 12 practice-area plugins for Claude, targeting the legal industry through Claude Cowork. The connectors link Claude to platforms like DocuSign, iManage, Relativity, Thomson Reuters CoCounsel, Harvey, and Everlaw, covering contract lifecycle, e-discovery, research, and deal rooms. The 12 plugins cover roles including Commercial Legal, Litigation, IP, Privacy, and AI Governance. Claude also integrates with Microsoft Word, Outlook, Excel, and PowerPoint, carrying context across all four apps.

Anthropic held its annual Code w/ Claude developer conference in San Francisco on May 12, 2026, with keynotes and breakout sessions now available on YouTube. The company announced doubled rate limits for Claude Code and raised API limits for Claude Opus. Four new Claude Managed Agents features launched: Dreaming (memory curation between sessions), multiagent orchestration with parallel subagents, Outcomes (a grader that improved task success by up to 10 points), and Webhooks. The conference continues in London and Tokyo.

Rivian has launched Rivian Assistant, an AI-powered voice assistant built directly into Gen 1 and Gen 2 vehicle hardware, rolling out via software update to Connect+ subscribers. Activated by saying "Hey Rivian" or holding the left steering wheel button, it handles hands-free vehicle controls, context-aware commands, navigation, messaging, and general knowledge queries. Built on Rivian Unified Intelligence, it features agentic integrations starting with Google Calendar, and includes privacy controls for data management.

AI IQ is a new ranking platform launched by Ryan Shea that scores frontier AI models using the human IQ scale. Rather than traditional leaderboard tables, AI IQ visualizes where models fall on the IQ bell curve, tracks how frontier AI intelligence changes over time, compares models on both IQ and EQ metrics, and displays what intelligence costs per unit. The platform aims to make AI capability comparisons more intuitive and accessible to a broader audience.

Krea has launched Krea 2, its first proprietary foundation model built entirely from scratch. Unlike adapted existing models, Krea 2 was designed specifically for aesthetic diversity and stylistic control in AI image generation. The release marks a significant shift for Krea, moving from tools built on third-party models to developing its own underlying technology. Early access is currently available for users interested in exploring its image generation capabilities.

Unitree has unveiled the GD01, a manned transformable mecha priced starting at $650,000. Described as the world's first production-ready manned mecha, the GD01 is classified as a civilian vehicle and weighs approximately 500 kilograms with an operator inside. The robot is capable of transforming, though specific transformation modes were not detailed. Unitree emphasized that users should operate the machine in a friendly and safe manner.

A South Korean team has developed a set of seven wireless AI-powered rings that translate American Sign Language and International Sign Language into text in real time. Worn on the seven most dominant signing fingers, the stretchy, accelerometer-equipped rings run on hour replaceable batteries and communicate via Bluetooth. Trained on 100 common signs, the system achieved over 88 percent accuracy even with first-time users, and includes AI autocomplete to predict words mid-sentence and keep pace with natural signing speeds.

Anthropic launched Agent View in Claude Code on May 11, 2026, a unified dashboard for managing multiple parallel Claude Code sessions from the CLI. Previously, developers had to juggle multiple terminal tabs and tmux grids. The new feature lets users kick off agents, send them to the background with `/bg` or `claude --bg [task]`, and reply to waiting sessions inline without switching context. Each row shows session status, last response, and interaction time. It is available as a Research Preview on Pro, Max, Team, Enterprise, and API plans.

OpenAI has launched Daybreak, a cybersecurity initiative combining frontier AI models with its Codex agentic platform to help defenders find, validate, and fix software vulnerabilities. The program offers three access tiers — standard GPT-5.5, GPT-5.5 with Trusted Access for Cyber, and GPT-5.Cyber — each with progressively more permissive capabilities for verified security workflows including red teaming and penetration testing. Cloudflare CTO Dane Knecht praised the platform for advancing agentic execution in security operations.

Google is expanding its AI-powered Google Finance to Europe, launching with full local language support. The revamped platform includes AI-powered research for stocks and market trends, a Deep Search feature now globally available, advanced charting tools with technical indicators like moving average envelopes, real-time news and expanded cryptocurrency and commodities data, and live earnings call coverage with synchronized transcripts and AI-generated insights including annotated highlights.

Google's Threat Intelligence Group (GTIG) reports that adversaries are using AI to develop zero-day exploits, build polymorphic malware, and run autonomous attack operations. For the first time, GTIG identified a cybercrime actor who used AI to discover and weaponize a 2FA bypass zero-day in a popular open-source admin tool, planning mass exploitation. China- and North Korea-linked groups are also using AI for vulnerability research, while Russia-nexus actors use it to generate obfuscation networks. AI malware like PROMPTSPY now autonomously interprets system states to execute commands.

Google has launched REPLIQA (Research Program at the Intersection of the Life Sciences and Quantum AI), committing $10 million through Google.org to five universities — Harvard, MIT, UC San Diego, UC Santa Barbara, and the University of Arizona. The initiative aims to apply quantum science and AI to biology, developing quantum sensors and algorithms to simulate complex molecular interactions, such as the P450 enzyme critical to drug development, with long-term goals of improving human health outcomes.

Anthropic has launched the Claude Platform on AWS, now generally available, giving AWS customers native access to the full Claude API with AWS IAM authentication, CloudTrail audit logging, and unified billing through a single AWS invoice. Unlike Claude on Amazon Bedrock, where AWS is the data processor, the new platform is operated by Anthropic and includes day-one access to new features like Claude Managed Agents, code execution, web search, Skills, and MCP connector.

OpenAI has launched the OpenAI Deployment Company, a new standalone business unit designed to embed specialized Forward Deployed Engineers into organizations to build and deploy AI systems. In connection with the launch, OpenAI agreed to acquire Tomoro, an applied AI consulting firm, bringing approximately 150 experienced engineers from day one. The venture launches with over $4 billion in initial investment and is backed by 19 partners including TPG, Bain Capital, Goldman Sachs, McKinsey, and SoftBank Corp.