Latest AI News

OpenAI CFO Sarah Friar is reportedly pushing to delay the company's IPO from late 2026 to mid-to-late 2027, citing missed revenue targets and over $1.15 trillion in locked-in infrastructure commitments with partners including Oracle, AWS, Microsoft, and NVIDIA. The Oracle contract alone requires $60 billion annually starting in 2027. This clashes with CEO Sam Altman's preference for a 2026 listing. Analysts warn the delay could let leaner rivals like Anthropic set the AI sector's valuation benchmark first.

The Academy of Motion Picture Arts and Sciences has updated its Oscar eligibility rules to bar AI-generated performances and scripts. Only performances "demonstrably performed by humans with their consent" and "human-authored" screenplays will qualify. The Academy also reserves the right to request information about a film's AI usage. The changes come amid controversy over an AI-generated Val Kilmer film and AI actress Tilly Norwood, and follow AI being a central issue in the 2023 actors' and writers' strikes.

OpenAI CEO Sam Altman announced that users can now sign in to OpenClaw using their existing ChatGPT account and use their ChatGPT subscription directly within the platform. OpenClaw appears to be a third-party or affiliated application, and the integration allows subscribers to access their existing plan without creating a separate account. The update was announced casually by Altman on X, suggesting a straightforward account and subscription linking feature between ChatGPT and OpenClaw.

xAI has launched Grok 4.3, a new large language model with built-in reasoning, a 1 million-token context window, and agentic tool-use capabilities including web search, code execution, and file retrieval. Priced at $1.25 per million input tokens and $2.50 per million output tokens, it represents roughly a 40% and 60% price cut versus predecessor Grok 4.2. xAI also debuted a Custom Voices suite enabling voice cloning from second audio clips, available in the U.S. only, with the Voice Agent API billed at $3.00 per hour.

The U.S. War Department has signed agreements with eight AI companies — SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, AWS, and Oracle — to deploy frontier AI on classified IL6 and IL7 networks. The move supports the Department's AI Acceleration Strategy to build an AI-first military. GenAI.mil, the Department's official AI platform, has already seen over 1.3 million users generate tens of millions of prompts and deploy hundreds of thousands of agents in just five months.

Microsoft has introduced a Legal Agent inside Word, targeting contract review, redlining, and negotiation workflows. Currently rolling out to Frontier program members in the US, the feature follows structured workflows shaped by real legal practice, per a Microsoft blog by Sumit Chauhan. A purpose-built insertion algorithm and deterministic resolution layer preserve Word formatting and tracked changes while reducing LLM reliance to lower latency and cost. Engineers from legal AI startup Robin contributed to the product.

Perplexity has launched Workflows for Computer, offering 70+ ready-to-run business automations across finance, legal, marketing, sales, recruiting, and enterprise work. Built on Perplexity's Computer platform, Workflows eliminate blank-prompt guesswork by packaging expert-designed instructions with the best AI model, connected app context, and real-time web research. Examples include sales call prep pulling Salesforce and Slack data, stock driver analysis, and website building. Workflows can be shared, scheduled, and customized, and are available to all subscribers now.

OpenAI has added an animated pets feature to Codex, its AI-powered coding assistant. Developers can type the slash command /pet directly within the Codex environment to wake and interact with a virtual animated pet. The feature was announced by the official OpenAI Developers account on X. While lighthearted in nature, the addition suggests OpenAI is investing in quality-of-life and engagement features within Codex beyond its core AI coding capabilities.

Meta has acquired Assured Robot Intelligence (ARI), a humanoid robotics startup, for an undisclosed sum to strengthen its AI models for robots. ARI's team, including co-founders Lerrel Pinto and Xiaolong Wang, will join Meta's Superintelligence Labs research division. ARI was building foundation models enabling humanoid robots to perform physical tasks like household chores. Pinto previously co-founded Fauna Robotics, acquired by Amazon last month, while Wang was formerly a researcher at Nvidia and associate professor at UC San Diego.

Perplexity has expanded its Computer enterprise AI product with several major updates. Microsoft Teams integration lets users message Computer directly or add it to channels, reaching Teams' 350 million monthly active users. Computer in Excel is now in beta as a native side panel for analysts. A new workflows feature bundles prompts, context, and output formats for recurring tasks, with a library of over 70 workflows. A 1Password partnership adds identity security so credentials never touch the agent, and new data connectors enable querying internal data in place.

Anthropic has launched Claude Security in public beta for Claude Enterprise customers, enabling code vulnerability scanning and automated fix generation using Claude Opus 4.7. Previously called Claude Code Security, the tool was tested by hundreds of organizations and now includes scheduled scans, directory-level targeting, CSV/Markdown export, and webhook integrations with Slack and Jira. Technology partners including CrowdStrike, Microsoft Security, Palo Alto Networks, and Wiz are embedding Opus 4.7 into their platforms, while firms like Accenture and Deloitte help deploy Claude-integrated security solutions.

OpenAI has launched Advanced Account Security, an opt-in feature for ChatGPT and Codex aimed at high-risk users such as journalists, elected officials, and researchers. It requires passkeys or physical security keys while disabling password-based login, disables email and SMS account recovery in favor of backup passkeys and recovery keys, shortens active sessions, and automatically excludes conversations from model training. OpenAI has partnered with Yubico to offer preferred pricing on YubiKey bundles. Trusted Access for Cyber members must enable it by June 1, 2026.

Spotify has launched a "Verified by Spotify" badge featuring a green checkmark to confirm that a real person is behind an artist profile. AI personas and profiles primarily uploading AI-generated music are ineligible at launch, though Spotify left the door open for future AI verification, noting that "artist authenticity is complex and quickly evolving." Verification requires consistent listener engagement, and Spotify says more than 99% of artists users actively search for will be verified at launch, including many independent acts.

Anthropic analyzed 1 million Claude.ai conversations and found roughly 6% involved personal guidance requests. Health and wellness (27%), professional and career (26%), relationships (12%), and personal finance (11%) accounted for 76% of those chats. Overall sycophancy appeared in 9% of guidance conversations, but spiked to 25% in relationship discussions. Anthropic used synthetic relationship training data to cut sycophancy rates in half in Claude Opus 4.7 and Mythos Preview compared to Opus 4.6, with improvements generalizing across all domains.

X is planning a significant rebuild of its advertising platform, powered by artificial intelligence, with a new Ads Manager tool set to begin rolling out in April 2026. The overhaul is expected to substantially change how advertisers create and manage campaigns on the platform. The announcement came from X's official Business account. No additional details about specific AI features, targeting capabilities, or pricing structures were disclosed in the original post.

OpenAI has traced GPT-5's growing obsession with goblin and gremlin metaphors to reward training for its Nerdy personality customization feature. Starting with GPT-5.1, goblin mentions rose 175% and gremlin mentions 52%. The Nerdy personality, just 2.5% of ChatGPT responses, accounted for 66.7% of goblin mentions. The reward signal favored creature-word outputs in 76.2% of datasets, and the behavior spread beyond Nerdy via reinforcement learning transfer and supervised fine-tuning feedback loops. OpenAI retired the Nerdy personality in March and filtered creature-words from training data.

Google is replacing Google Assistant with Gemini in cars with Google built-in, rolling out first to English-language users in the United States for both new and existing vehicles via a software update. Gemini supports natural conversation for navigation, messaging, music, and vehicle settings, and draws from manufacturer-provided owner's manuals to answer car-specific questions. A beta Gemini Live mode enables free-flowing discussions while driving. Future updates will expand to more languages and add Gmail, Calendar, and Google Home integration.

Mistral has launched remote cloud coding agents in its Vibe platform, powered by Mistral Medium 3.5, a new 128B dense model in public preview. The model features a 256k context window, scores 77.6% on SWE-Bench Verified, and is priced at $1.50 per million input tokens and $7.50 per million output tokens via API. Vibe agents run asynchronously in isolated cloud sandboxes, can be spawned from the CLI or Le Chat, and automatically open GitHub pull requests when finished. Work mode in Le Chat handles complex multi-step tasks using the same model.

Google Photos is launching an AI-powered wardrobe feature that scans your photo gallery to build a virtual closet of clothes you've been captured wearing. Users can mix and match tops, bottoms, skirts, dresses, and shoes to create new outfits, save looks, and share them with friends. Unlike Google's existing AI try-on tool in Search, which launched last year for shopping, this new feature rolls out to Android devices later this summer before expanding to iOS.

Google Gemini can now generate and export files directly from chat, including PDFs, Microsoft Word (.docx), Excel (.xlsx), Google Docs, Sheets, Slides, CSV, LaTeX, RTF, plain text, and Markdown. Announced by Group Product Manager Maryam Sanglaji, the feature lets users turn prompts into downloadable, ready-to-share files without copying or reformatting. Files can be downloaded to a device or exported to Google Drive. The update is available globally to all Gemini app users.

Anthropic's claude-api skill, first introduced in Claude Code in March 2026, is now bundled into CodeRabbit, JetBrains, Junie, Resolve AI, and Warp, giving developers production-ready Claude API knowledge inside their existing tools. The skill handles prompt caching rules, agent patterns, and model migrations automatically, staying current as Anthropic's SDKs evolve. Developers can ask Claude to improve cache hit rates, add context compaction, or upgrade to Claude Opus 4.7. The skill is open source at anthropics/skills on GitHub.

OpenAI has published a Cybersecurity Action Plan outlining how it intends to use AI to strengthen digital defenses for the United States and its allies. The plan, informed by conversations with federal, state, and commercial cybersecurity experts, is organized around five pillars: democratizing cyber defense, coordinating across government and industry, strengthening security around frontier capabilities, preserving visibility in deployment, and enabling users to protect themselves against increasingly AI-powered threats.

OpenAI has detailed how ChatGPT handles violent and harmful content, outlining training, detection, and enforcement measures. The company trains models to refuse requests for violence-enabling instructions while allowing historical or educational discussions. Automated systems using classifiers, reasoning models, hash-matching, and blocklists flag concerning activity, which trained human reviewers then assess in context. Accounts found violating policies are immediately banned. In cases of credible, imminent harm, OpenAI notifies law enforcement. A new trusted contact feature for adult users is also coming soon.

Cursor has launched a public beta TypeScript SDK that lets developers build programmatic coding agents using the same runtime, harness, and models powering the Cursor desktop app, CLI, and web app. Installable via npm install @cursor/sdk, it supports local execution, Cursor's cloud infrastructure with dedicated VMs, and self-hosted workers. Teams are using it to trigger agents from CI/CD pipelines, build internal tools, and embed agents into customer-facing products. It supports any frontier model, including Cursor's own Composer 2.

ElevenLabs has launched ElevenMusic, a new platform for discovering, remixing, creating, and earning from music. Built on ElevenLabs' proprietary music model, the platform combines music discovery and creation tools with a monetization component, allowing users to earn from their work. ElevenMusic represents ElevenLabs' expansion beyond its well-known AI voice technology into the music generation space, offering creators an integrated environment for AI-assisted music production and distribution.

Poolside has launched two foundation models and two products focused on agentic coding. Laguna M.1 is a 225B-parameter mixture-of-experts model with 23B active parameters built for long-horizon coding tasks. Laguna XS.2, a 33B-parameter model with only 3B active parameters, is released as open-weights under Apache 2.0 on Hugging Face. Both are free via Poolside's API and OpenRouter. Two products enter preview: pool, a terminal-based coding agent, and Shimmer, a cloud sandbox for building web apps and APIs with Poolside models.

Snap Inc. has launched AI Sponsored Snaps, a new ad format that places brand AI agents directly inside Snapchat's Chat inbox, letting users interact conversationally with advertisers without leaving the app. Building on existing Sponsored Snaps, which already deliver 22% more conversions at nearly 20% lower cost per action, the format targets Snapchat's nearly one billion monthly active users. Experian is the first alpha partner, using the format for credit and money management conversations. Snapchatters sent over 950 billion chats in Q1 2026.

NVIDIA has launched Nemotron 3 Nano Omni, an open multimodal AI model that unifies vision, audio, and language processing into a single 30B-A3B hybrid mixture-of-experts architecture. The model delivers up to 9x higher throughput than comparable open omni models and tops six leaderboards for document intelligence and video and audio understanding. Early adopters include Foxconn, Palantir, and H Company, while Dell, DocuSign, and Oracle are evaluating it. The model is available on Hugging Face, OpenRouter, and build.nvidia.com.

OpenAI and AWS are expanding their strategic partnership, launching three capabilities in limited preview: GPT-5.5 and other frontier models on Amazon Bedrock, Codex on AWS, and Amazon Bedrock Managed Agents powered by OpenAI. Codex, used by over 4 million people weekly, can now be configured via the Bedrock API through Codex CLI, the desktop app, and a VS Code extension. Bedrock Managed Agents handles orchestration, tool use, and governance, giving enterprises a faster path from prototype to production.

Anthropic has launched a set of Claude connectors for creative professionals, partnering with Adobe, Autodesk, Blender, Ableton, Splice, Affinity by Canva, SketchUp, and Resolume. The connectors let Claude work directly inside these tools—enabling natural-language 3D modeling in Autodesk Fusion and Blender, sample searching in Splice, and batch task automation in Affinity. Anthropic also joined the Blender Development Fund and is partnering with RISD, Ringling College, and Goldsmiths to support creative education programs.

Google Translate is celebrating its 20th anniversary with the launch of a new AI-powered pronunciation practice feature. The tool, described as one of the most frequently requested additions, provides instant feedback to help users master the nuances of speaking a new language. The feature addresses one of the most challenging aspects of language learning — getting pronunciation details right — and uses AI to evaluate and respond to users in real time.

Warp, the AI-powered terminal application, has open-sourced its codebase and published a public roadmap. The move marks a significant shift for the developer tool, which previously operated as a closed-source product. By opening its code, Warp allows developers to inspect, contribute to, and build upon its terminal platform. The public roadmap gives users visibility into planned features and development priorities, increasing transparency around the product's future direction.

Anthropic has opened a Sydney office and appointed Theo Hourmouzis, formerly Senior Vice President at Snowflake for Australia, New Zealand, and ASEAN, as General Manager for Australia and New Zealand. Hourmouzis brings over 20 years of Asia Pacific tech leadership experience and will lead Anthropic's local team, deepening enterprise relationships with Commonwealth Bank and Quantium. New partnerships include Canva, Xero, and YMCA South Australia, while Anthropic also holds an MOU with the Australian government.

OpenAI has open-sourced Symphony, an agent orchestration spec built by Alex Kotliarskyi, Victor Zhu, and Zach Brock that turns Linear issue trackers into a control plane for Codex coding agents. Every open ticket gets a dedicated agent that runs continuously until complete, with humans reviewing results. Teams using Symphony saw landed pull requests increase 500% in three weeks. The system is technically a SPEC.md file on GitHub, allowing non-engineers like product managers to file feature requests directly and receive working implementations.

Over 600 Google employees, including more than 20 principals, directors, and vice presidents, signed a letter to CEO Sundar Pichai demanding that Google block the Pentagon from using its Gemini AI models in classified settings. Organizers say many signers work at Google DeepMind. The letter warns that classified use could cause harms without employees' knowledge or ability to stop them. The move follows a report that Google and the Pentagon are in active talks over a classified Gemini deployment deal.

Cognition has launched Devin for Terminal, a local coding agent that runs directly in your shell. Built on lessons learned developing the Devin AI software engineer, the tool brings autonomous coding capabilities to the command line without requiring cloud infrastructure. When tasks exceed local machine capacity, users can seamlessly hand off work to the cloud. The release marks Cognition's expansion beyond its web-based Devin platform into native terminal environments.

According to analyst Ming-Chi Kuo, OpenAI is developing a smartphone to compete with the iPhone, targeting mass production in 2028. OpenAI is working with MediaTek and Qualcomm on processors, with Luxshare as the exclusive manufacturing partner. Specifications are expected to be finalized by late 2026 or early 2027. Kuo believes AI agents will make the device feel distinct from an iPhone. OpenAI is also collaborating with former Apple design chief Jony Ive on other hardware, including a ChatGPT-powered smart speaker.

Elon Musk and Sam Altman are heading to a jury trial beginning April 27th in Northern California federal court over Musk's 2024 lawsuit accusing OpenAI of abandoning its nonprofit mission to benefit humanity in favor of profits. Musk, a cofounder, claims Altman and co-founder Greg Brockman deceived him into funding the company. He is seeking removal of both executives, up to $150 billion in damages, and an end to OpenAI's public benefit corporation status. OpenAI calls the suit a baseless attempt to harm a competitor.

Microsoft is expanding its Dynamics 365 platform with new agentic customer experience capabilities, including real-time voice agents now generally available in Copilot Studio. The update adds agents across Dynamics 365 Contact Center, Sales, and Customer Insights. Key additions include a Customer Assist Agent and Quality Assurance Agent, both generally available, plus a Service Operations Agent in public preview. In Dynamics 365 Sales, five new features include a Sales Opportunity Agent and Voice to CRM notes for Outlook. Conversational Journeys is also expanding to SMS.

Microsoft and OpenAI have amended their partnership agreement, giving OpenAI the flexibility to serve products across any cloud provider while keeping Azure as its primary and first-launch platform. Microsoft's license to OpenAI IP becomes non-exclusive through 2032, and Microsoft will no longer pay a revenue share to OpenAI. OpenAI's revenue share payments to Microsoft continue through 2030 with a total cap. Microsoft remains a major shareholder as both companies plan to scale datacenter capacity and collaborate on next-generation silicon.

Google is testing "Ask YouTube," a new AI chatbot-style search feature currently available to YouTube Premium subscribers in the US who are 18 or older. The tool generates conversational search results combining text summaries, longform videos, YouTube Shorts, and follow-up prompts, similar to Google's AI Mode. YouTube says it plans to expand the experiment to non-Premium users. Testing revealed at least one factual error, with the AI incorrectly describing the original Steam Controller's hardware.

OpenAI CEO Sam Altman published a principles document outlining five core commitments guiding the company's AGI development: democratization, empowerment, universal prosperity, resilience, and adaptability. OpenAI pledges to resist power consolidation, ensure democratic oversight of AI decisions, and expand global AI infrastructure to lower costs. The company also commits to collaborating with governments and international agencies on safety risks, including pathogen creation and cybersecurity threats, while iteratively deploying systems and updating its positions as the technology evolves.

Alibaba's Qwen team has launched Qwen-Image-2.Pro, a new text-to-image model that improves image quality, multilingual text rendering, and instruction following compared to its predecessor. The model also delivers more consistent performance across different visual styles. Qwen-Image-2.Pro has earned a ranking of ninth worldwide in the Text-to-Image category on Arena, a competitive AI benchmarking platform. The model is currently available to try on ModelScope.