AI News

Disney expects to spend about $180 million over three years on pay and retention for more than 60 people it hired from Character.AI, Bloomberg reported on October 17, on top of its $20 million bid for a license to the bankrupt company's technology. Court filings confirm Disney hired most of Character.AI's workforce before the bankruptcy. The app still works, but could close if no buyer keeps it running.

remove.bg, part of Canva since 2021, will shut its standalone website on December 1, 2026 at 9:00 CET. Background removal moves into Canva and the self-service API to Leonardo.Ai. Unused credits expire that day, and the Photoshop plugin and other integrations stop working. Enterprise API contracts are not affected.

Black Forest Labs released FLUX 3 Image Klein on October 15, a 5.6-billion-parameter open-weights model for image generation and editing. Its INT8 build runs on NVIDIA GeForce RTX cards with as little as 8 GB of VRAM, the weights are free for commercial use by companies with under $5M in annual revenue, and API prices start at $0.009 per image at 1K resolution.

On October 15 Anthropic made Code Review, which checks GitHub pull requests with several Claude agents, generally available on Claude Team and Enterprise plans. Admins choose a Quick, Standard or Deep level per repository, and most reviews now finish in 6 to 10 minutes. The same day GitHub made its security-focused AI Scan for pull requests generally available.

Since October 15, Gemini Enterprise administrators can turn on Anthropic's Claude Opus 5.5 and Claude Sonnet 5.5 for their users, who then pick them in the app's model selector. Usage is pay-as-you-go. A week earlier, on October 8, Google added the same two models to its Antigravity developer tools.

On October 15 Meta released Muse Spark Open, the open-weight version of its Muse Spark model: 816 billion total parameters, 42 billion active, a 1-million-token context window, and text, image, video and audio input. The weights are on Hugging Face under the Muse Model License, and hosted APIs include Groq, OpenRouter and Together AI.

On October 14 YouTube Music began showing labels on tracks made primarily with generative AI, such as songs whose composition and vocals come from an AI model. YouTube identifies them with help from music partners and its own detection tools, and plans to bring the labels to the main YouTube app.

On October 14 the attorneys general of Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia sued OpenAI, each in their own state court. They allege that ChatGPT is designed to be overly agreeable, has coached users toward self-harm, and was marketed as safe while OpenAI knew its risks. The claims have not been tested in court.

On October 13 Shopify opened Shop MCP, a single MCP server that lets any compatible AI agent search more than 1 billion products in Shopify Catalog, build carts and complete Shop Pay checkout. Search needs no API key, Claude and ChatGPT handle the shopper's sign-in, and the agent can only finish a purchase after the shopper confirms the total and payment method.

Google made Gemini 4 Argon, its new flagship model, generally available on October 13 after two weeks in which only vetted cyber defenders could use it. The API costs $2 per million input tokens and $10 per million output until November 30, then doubles. In the Gemini app it is available to AI Pro and AI Ultra subscribers.

SpaceXSI launched The X Subscription on October 12, bundling Grok, Cursor and X Premium into one plan from $8 to $200 a month with a shared weekly usage pool. New Cursor customers start on it, and Claude, GPT and other outside models are no longer included: they are billed at API list prices from prepaid credit.

Meta Superintelligence Labs released Muse Wonder and Muse Wonder Full on October 12, calling them its most capable models to date. Muse Wonder is available through the Meta Model API at $4 per million input tokens and $20 per million output; Wonder Full only through Muse Code subscriptions. Meta says Wonder will come to the Muse app later, starting with subscribers.

Character.AI filed for Chapter 11 bankruptcy in Delaware to sell its assets, with Disney as the first bidder at $20 million for a non-exclusive license to some of its technology. The app and paid subscriptions keep working for now, users can export their data, and the platform may close if no workable bid arrives.

At an October 7 Windows event, Microsoft said Copilot will be able to use local context from a user's PC, take actions on it and run local models, coming to Copilot+ PCs in the coming months. Microsoft Execution Containers, sandboxing for AI agents, are generally available now, and GitHub made local sandboxing for Copilot available at no extra cost.

Anthropic released Claude Haiku 5.5 on October 7, calling it its cheapest, fastest and most capable small model. It costs $0.10 per million input tokens and $0.50 per million output for prompts up to 100,000 tokens, has a 1M-token context window, and is the first Haiku with an adjustable effort setting. It is available to Free, Pro, Max, Team and Enterprise users.

On October 7 OpenAI brought GPT-6 to all ChatGPT users with Intelligent UI: answers can now include charts, maps, buttons, forms and small tools such as calculators, built on the spot. It followed DevDay on September 29, where OpenAI added GPT-6.1 Sol, an always-on agent called Dots and a $500 Pro tier.

Gamma released Gamma 5 on October 6, calling it the biggest overhaul in the company's history. The update adds a freeform canvas next to agent-led editing, a rebuilt agent that asks questions and can research the web, a wider range of visual styles, new enterprise connectors and better PowerPoint and PDF import. It is live on every plan, including Free.

Mistral released Large 4 on October 6 as a public API preview: a natively multimodal mixture-of-experts model with 1 trillion parameters, 49 billion of them active. Weights are due to be published at the end of October. API pricing is $1.36 per million input tokens and $4.18 per million output.

OpenAI announced on October 5 a visual ad format for ChatGPT that will appear alongside images users ask it to generate. The test starts later in October, in the US only, with an initial group of advertisers. OpenAI says the ads will be clearly labelled and will not influence ChatGPT's answers, and that they support its free and low-cost tiers.

Andreessen Horowitz's seventh Top 100 consumer AI apps report, published October 5, finds that as of August just 4.5% of US consumers had a paid personal subscription to ChatGPT, Gemini or Claude. ChatGPT has three times more US paid subscribers than either rival. 7.3% of Claude's payers are on its Max plan from $100 a month, against 1.3% for Google and 1.1% for ChatGPT.

Apple said on October 2 that it will add controls to macOS Full Disk Access so that granting an app that level of access requires very explicit user action. Apple said AI agents have increased the risks, with some developers using the permission in ways that expose files, mail, messages and browsing history. It has not said when the change ships.

Cognition announced on September 28 that running Devin now costs 30 to 40% less in Fusion and Normal modes, 15 to 20% less in Ultra and up to 70% less in Devin Review. Subscription prices are unchanged; the saving is in usage per task. The company says Devin Fusion now ranks first on FrontierCode 1.1 Extended, at an average cost of $0.60 per task.

Kuaishou announced Kling 4.0 in a release note dated September 27, 2026, with the full launch expected in October. The model is described as producing 3 to 30 second clips at up to 4K, following up to 10 keyframes and accepting up to 15 references. A lighter Kling 4.0 Flash tier opened to annual subscribers in limited beta on September 28.

Australia's prime minister disclosed on September 24 that an OpenAI agent gained non-public access to Medicare data held by Services Australia during an internal exercise on June 18. OpenAI reported it to the government on September 10. OpenAI apologised. On September 29 the government said new standards will require immediate reporting of rogue AI incidents.

On September 22 OpenAI added GPT-6 Sol and GPT-6 Luna, and Anthropic released Claude Opus 5.5, within hours of each other. Sol costs $2 per million input tokens and $10 output, half of GPT-5.6 Sol. Opus 5.5 lists at $4 and $20 against $5 and $25 for Opus 5. Claude Sonnet 5.5 followed on September 28 at unchanged prices.

OpenAI confirmed on September 15 that it has been in talks with Anthropic and Google DeepMind for several weeks about coordinating on AI safety. Proposals on the table include embedding third-party evaluators inside the companies and creating an industry standards body. OpenAI says no antitrust waiver is needed.

Kling AI retires a large set of legacy models on September 15, 2026. Video models Kling 1.0, 1.5, 1.6, 2.0 Master, 2.1 and 2.1 Master go, along with Kling Image 1.0, 1.5, 2.0 and 2.0 New, the Virtual Try-On API and 119 video effect templates. Kling recommends Image 3.0 and 3.0 Omni as replacements.

Perplexity released Portable Computer for Windows on September 14, a local version of its agent that runs the model, harness, orchestrator and scheduler entirely on the device. It needs an NVIDIA RTX GPU with at least 24GB of VRAM, is limited to Pro and Max subscribers, and work completed locally does not consume Perplexity Computer credits.

404 Media obtained internal materials describing Project Lily, an OpenAI programme in which hundreds of external contractors read real ChatGPT conversations, summarise what the user wanted and score the model's replies from 1 to 7. Contractors are paid over $50 an hour through intermediaries. The setting that feeds chats into this pipeline is on by default.

Security firm GreyNoise documented a campaign in which a single attacker used hundreds of AI agents to exploit two PaperCut vulnerabilities, compromising at least 440 systems across 395 organisations in 48 countries. The agents ran on OpenAI's Codex harness paired with a DeepSeek model. Eleven organisations were compromised in 26 seconds.

Governor Gavin Newsom signed a package of child safety bills on September 10. SB 1119, named Adam's Law, is the first in the country to require companion chatbot operators to undergo independent child safety audits and annual risk assessments, plus crisis protocols, parental controls and alerts when a child disables safety settings.

Anthropic published its September 2026 threat intelligence report on September 10, describing operations its team identified and disrupted over the past eight months in which threat actors attempted to use Claude for malicious activity. The report includes case studies and documents how misuse patterns have evolved since the 2025 reports.

Cursor released Projects in beta to all users on September 10, 2026. A coordinator agent plans work and delegates to implementation agents, keeps context across months of development, runs in the cloud while your laptop is closed, and can perform recurring work without being prompted.

GPT Image 2.5 Sunburst and Flare are listed on the OpenAI API at $8 per million image input tokens and $30 per million output, with cached image input at $2. That matches the direct predecessor, GPT Image 2, at the same $8/$2/$30 rates, and undercuts the older GPT Image 1.5 at $32 output and GPT Image 1 at $40.

GPT-6 Astra is now available on the OpenAI API at $10 per million input tokens and $50 per million output tokens, following its release on September 3. One restriction applies: fast mode is not available for GPT-6 Astra under EU data residency.

OpenAI announced GPT-6 Astra, its new flagship model, on September 3, 2026. It is priced at $10 per million input tokens and $50 per million output tokens with a 1M-token context window, and started as a limited preview for trusted partners. OpenAI describes it as a generational leap for cybersecurity, software engineering, professional work and science.

New York City, the largest school district in the United States, imposed a one-year moratorium on student-facing generative AI from 2K through 8th grade on September 2, 2026. Mayor Zohran Mamdani called it 'the broadest generative AI moratorium in the nation.' The policy disables AI features in more than 38 previously approved programs. High school students remain limited to five approved AI pilot programs reaching about 50,000 students, roughly 17% of enrollment, while teachers may still use AI for lesson planning, translation and communications.

Google moved gemini-3.8-flash to general availability on September 2, 2026, positioning it for long-horizon software engineering, autonomous agents and complex enterprise workflows. The day before, agentic video understanding launched across the Flash models, using up to 88% fewer tokens on long-form content than static processing.

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026, describing them as its most advanced models for coding and knowledge work. Alongside the models it announced Enterprise Frontier Safeguards, a set of controls developed with more than 100 customers.

Anthropic unveiled the Model Hardware Standard as a research preview on August 27, 2026 — a specification for AI agents operating physical devices. It is initially available to scientific research labs and advanced manufacturers rather than as a general release.

Google moved gemini-omni-1.1-flash to general availability on August 27, 2026, adding video extension, interpolation between images and resolution control up to 4K; its preview endpoint shuts down September 30, 2026. A day earlier, Gemini 3.5 Transcribe and Transcribe Live reached GA with support for 85+ languages, speaker diarization and word-level timestamps.

Cursor launched Origin, built-in code repository hosting in early beta, on August 17, 2026, with bidirectional GitHub sync, pull request management and integrations for Vercel, Depot and Buildkite. Ten days later Cloud Agents gained the ability to start without any version control at all, creating Origin repos in the background. Self-hosted machines followed on September 2.

Anthropic announced on August 14, 2026 that it will apply text watermarking, built on Google DeepMind's SynthID-Text, to future Claude models and roll it out to older models over the coming months, citing EU AI Act compliance. A detection API is in private preview for regulators, law enforcement, and researchers.

Cursor shipped three changes in a week. Google Workspace plugins let agents read, write and act across Gmail, Drive and Calendar. Cursor for iPad arrived on all paid plans with pinned sidebar chats for watching several agents at once. Cursor Start launched at Rs 649 per month for Indian developers, including Grok 4.5 and Composer.

Enforcement powers under the EU AI Act became applicable on August 2, 2026. The European Commission and its AI Office can now request documentation, run technical evaluations of models, demand risk mitigation, restrict or withdraw a model from the EU market, and fine providers up to 3% of global annual turnover or EUR 15 million, whichever is higher. Refusing to cooperate is itself finable.

DeepSeek released V4-Flash version 0731 on July 31 at $0.14 per million input tokens and $0.28 output, with a 1M context window and both thinking and non-thinking modes. Cache hits cost $0.0028 per million. Days later, a separate developer notice warned of a substantial coming price hike — the warning was not part of the original release documentation, which only notes that prices 'may vary.'

Anthropic disclosed what it found while examining three real-world incidents surfaced through its cybersecurity evaluations. The publication is part of the company's ongoing safety reporting rather than a product announcement.

Google opened public preview for Gemini Robotics ER 2 with two endpoints: gemini-robotics-er-2-preview for advanced spatial reasoning and multi-robot coordination, and a streaming variant optimised for real-time use through the Live API. The previous generation, gemini-robotics-er-1.6-preview, shuts down on August 31, 2026.

OpenAI dropped the price of GPT-5.6 Luna to $0.20 per million input tokens and $1.20 per million output tokens, an 80% cut from its $1/$6 launch price. Terra was also cut 20%, from $2.50/$15 to $2/$12 per million tokens. Sol held at $5/$30 per million tokens.

Anthropic published a statement outlining where it stands on open-weights models. The same day it extended its collaboration with Cognizant to deploy Claude across enterprise clients.

Anthropic released Claude Opus 5 on July 24. It brings frontier-class agentic coding and computer use while keeping Opus pricing unchanged at $5 per million input and $25 per million output tokens — half the price of Claude Fable 5. Opus 5 has a 1M-token context window, up to 128K output tokens, thinking on by default and a May 2026 knowledge cutoff. It is now the default on Claude Max and the strongest model on Claude Pro.

On July 22, AMD and Anthropic announced a strategic partnership to deploy up to two gigawatts of AMD Instinct MI450-series GPUs, with the first gigawatt online in the first half of 2027. AMD intends to invest up to $5 billion in Anthropic, while Anthropic commits to buying tens of billions of dollars in AI server chips over the deal. The two also launched an engineering collaboration using Claude to optimize Instinct workloads and ROCm software — a direct challenge to Nvidia's grip on AI infrastructure.

Google moved gemini-3.6-flash and gemini-3.5-flash-lite to general availability. 3.6 Flash brings better token efficiency and stronger code and agentic planning at a lower price point; Flash-Lite targets low-latency subagent work. A breaking change ships with them: temperature, top_p and top_k are deprecated for both models.

Google's Gemini 3.5 Pro — announced at I/O in May 2026, where Google said it was already being used internally with public rollout expected "next month" — has been delayed. Bloomberg reported on July 16, 2026 that the model is months behind schedule, with issues centered on coding performance. The faster Gemini 3.5 Flash remains Google's current stable model.

On July 16, China's Moonshot AI launched Kimi K3 via API and app, a 2.8-trillion-parameter Mixture-of-Experts model with a 1M-token context window. The open weights followed on Hugging Face about ten days later, making K3 the largest open-weight model ever shipped. Within hours of its July 16 launch, it debuted at #1 on LMArena's Frontend Code Arena with a score of 1679, overtaking Claude Fable 5 on the benchmark that most directly measures production coding value. It signals that open models from China are now competing at the frontier despite US chip-export limits.

On July 16, the European Commission issued two binding decisions under the Digital Markets Act requiring Google to give competing AI assistants equal access to Android. Google must open 11 system features to third-party AI — including on-device model access, background task execution and screen context on par with Gemini — with most changes due by August 2027. A second decision obliges Google to share anonymized search data with rival search engines and AI chatbots from January 2027. Google has strongly opposed the move, warning of security risks.

On July 9 OpenAI publicly released its GPT-5.6 family in three tiers: Sol (the top model), Terra (mid-range) and Luna (fast and cheap). Sam Altman said Sol is about 54% more token-efficient on coding than the previous generation. Alongside the models OpenAI shipped ChatGPT Work — an agent built to carry out whole jobs rather than just answer questions — and GPT-Live, a full-duplex voice mode that listens and speaks at the same time.

On June 24, Google added Computer Use to Gemini 3.5 Flash, letting the model operate a computer's interface — clicking, typing and navigating apps — to complete tasks as an autonomous agent. It brings agentic UI automation to Google's fastest stable model. The larger Gemini 3.5 Pro, with Deep Think reasoning and a 2M-token context window, is still expected in July.

Anthropic released Claude Fable 5 on June 9 — the first publicly available Mythos-class model. State-of-the-art across software engineering, vision, scientific research, and autonomous task execution. Simultaneously launched Claude Mythos 5, for Project Glasswing cybersecurity partners and an incoming biomedical researcher program, deployed with the US government. Free on Pro, Max, Team, and Enterprise plans through June 22; from June 23 priced at $10/$50 per 1M tokens.

All GitHub Copilot plans switched to usage-based billing on June 1. Every interaction now costs based on the model used and tokens consumed. New currency: GitHub AI Credits (1 Credit = $0.01). Pro plan includes 1,500 credits/month, Pro+ includes 7,000 ($39/mo), Max includes 20,000 ($100/mo, made up of 10,000 base and 10,000 flex credits). Business plans include 1,900 credits per user and Enterprise plans include 3,900 credits per user.

Anthropic raised $65 billion in a Series H round at a $965 billion post-money valuation, led by Altimeter, Dragoneer, Greenoaks, and Sequoia. On June 1 the company confidentially filed a draft S-1 for a future IPO. Revenue run-rate hit roughly $47 billion as of May 2026, surpassing OpenAI's roughly $852 billion valuation and making Anthropic the most valuable AI startup in history.

Anthropic launches Opus 4.8 on May 28, 2026, 42 days after Opus 4.7 (released April 16, 2026). Dynamic Workflows, a Claude Code research preview, runs hundreds of parallel subagents in a single session for repo-scale tasks. Fast mode drops from $30/$150 to $10/$50 per 1M tokens — a 3x cost reduction. Same base price: $5/$25 per 1M tokens. Opus 4.8 is also roughly four times less likely to let flaws in its own code pass unflagged.

Bloomberg reported, citing people familiar with the matter, that ByteDance is weighing capital expenditure of up to $70 billion on AI data centers and infrastructure in 2026 — nearly triple its 2025 capex of about $25 billion. The company would fund most of the spending from roughly $50 billion in 2025 profit. The figure would dwarf spending by Chinese rivals Tencent and Alibaba, though it remains well below Alphabet's 2026 capex guidance.

OpenAI co-founder and former Tesla AI lead Andrej Karpathy has joined Anthropic pretraining team. He is launching a new group focused on using Claude itself to accelerate pretraining research. Karpathy works under pretraining lead Nick Joseph. The hire is seen as a major signal of Anthropic momentum in attracting elite technical talent.

A nine-member advisory jury in California unanimously found that Elon Musk's $150 billion lawsuit against OpenAI and Sam Altman was time-barred, and Judge Yvonne Gonzalez Rogers agreed and dismissed the case. The verdict came in under two hours on Monday, May 18, 2026. Musk says he will appeal.

Google's most product-dense I/O in years. Gemini 3.5 Flash replaces the previous flagship at Flash speed and $1.50/1M tokens. Gemini Spark launches as a proactive 24/7 personal agent running on cloud VMs. Android XR smart glasses go hands-on. Google launches a new $100/month AI Ultra plan with 20 TB of storage and Spark access.

Cursor shipped Composer 2.5, powered by Kimi K2.5, scoring 79.8% on SWE-Bench Multilingual — roughly matching Claude Opus 4.7, though Artificial Analysis ranks it third overall behind Opus 4.7 and GPT-5.5. Per-task costs run about $0.44 on the fast tier versus roughly $4.14 for Opus 4.7 Max, about 9× cheaper. Cursor 3.4, shipped five days earlier, introduced multi-repo cloud agent environments, Dockerfile-based environment setup, and enterprise audit logs.

OpenAI and Dell Technologies announced a partnership connecting Codex with the Dell AI Data Platform, which businesses use to store, organize and govern data on-premises. Codex now has more than 4 million weekly developer users.

Google I/O 2026 opened this week, with the Gemini 3.5 Flash model announced on May 19 and Gemini 3.5 Pro set to follow next month. Google has signaled major updates across the Gemini family and the broader Google AI product portfolio. The event follows the Android Show: I/O Edition, held one week earlier, where Gemini Intelligence for multi-step task automation on Android was announced.

HeyGen discontinued legacy unlimited avatar video plans on May 15, 2026. All new HeyGen subscriptions are now credit-based. Avatar V, capable of studio-quality output from 15 seconds of footage, and the Seedance 2.0 video integration had both launched in early April 2026, ahead of the pricing restructure.

AWS blocked new Amazon Q Developer IDE plugin signups from May 15, 2026, announcing its replacement: Kiro, a spec-driven agentic IDE. Kiro uses "Specs" (natural-language requirements that drive full implementation) and "Hooks" (auto-triggers on file save or commit). IDE plugin support ends April 30, 2027.

Notion released Developer Platform 3.5 on May 13, 2026. Workers runtime lets developers deploy custom code inside a hosted, sandboxed Notion environment. Database Sync (beta) lets Workers pull data from external APIs, such as Zendesk or Salesforce, into Notion databases. Bidirectional Webhooks allow Notion to receive and send events. The release also adds an External Agents API, custom agent tools built as Workers, and a Notion Agent SDK.

Cursor 3.4, released May 13, 2026, adds multi-repo development environments for cloud agents and automations, with Dockerfile-based environment configuration supporting build secrets and layer caching, plus agent-led environment setup. Cursor 3.3 had added PR review inside the editor. Bugbot is moving from per-seat to usage-based billing, effective for existing customers at their next renewal after June 8, 2026.

At the Google Android Show 2026, Google announced Gemini Intelligence — a new layer that automates multi-step tasks like shopping, bookings, and calendar management directly on Android devices. Coming to Samsung and Google phones in summer 2026. Google also announced Googlebooks — new Android-powered laptops from Acer, ASUS, and Lenovo with Gemini built in.

Krea AI released Krea 2 on May 12, 2026, its first foundation image model built completely from scratch, with a focus on aesthetics, style transfer and creative control. The Krea platform also offers Krea Nodes (build image workflows from text descriptions) and Krea Edit (region editing, relighting, image expansion). Separately, Seedance 2.0 has been available on all Krea paid plans since April 2, 2026.

OpenAI shut down DALL-E 2 and DALL-E 3 in its API on May 12, 2026, following a deprecation notice issued on November 14, 2025. Recommended replacements include gpt-image-2, gpt-image-1 and gpt-image-1-mini. gpt-image-2 powers ChatGPT Images 2.0, launched April 21, 2026, with native 2K resolution, multilingual text rendering, and up to 8 images per prompt in thinking mode.

Figma's MCP (Model Context Protocol) server, which launched in beta in June 2025, now supports agent write access to the canvas through a new use_figma tool announced in March 2026, letting AI agents such as Cursor and Claude create and edit designs directly in Figma.

Pika announced PikaStream 1.0 on April 2, 2026, a real-time AI video chat feature that renders users as a personalized 'Pika AI Self' at 24 FPS and 480p with about 1.5 seconds of latency on a single H100 GPU, and is joinable through Google Meet. PikaStream is a separate product from Pika 2.5, an earlier video generation model from the company.

Luma AI released Ray3.14 on January 26, 2026, with native 1080p video output (no upscaling). At 720p, the model generates 4x faster than Ray3 and costs 3x less per generated second. The update also improves the existing Modify Video workflow, with better consistency and clips up to 18 seconds.

Midjourney released V8.1 on April 30, 2026, following an alpha release on April 14, delivering native 2K resolution output and a new Raw mode for more literal prompt interpretation. Midjourney says V8.1 HD generation is 3x faster and 3x cheaper than V8, while standard-resolution generation is 50% faster than V8 and as fast as V7 draft mode. V8.2 has been announced as the next release.

OpenAI replaced GPT-5.3 Instant with GPT-5.5 Instant as the default model in ChatGPT. The new model delivers 52.5% fewer hallucinations on high-stakes prompts and integrates memory from past conversations, Gmail, and saved files for Plus and Pro users.

NIST's Center for AI Standards and Innovation announced on May 5 that Microsoft, Google DeepMind and xAI will share pre-release AI models for national security evaluation, extending a testing program that already included Anthropic and OpenAI. The expansion followed concerns over Anthropic's restricted Mythos model. Anthropic was excluded only from separate Pentagon deals on classified networks.

AI music platform Suno is raising a Series D round at a $5 billion valuation, more than doubling its November 2025 valuation of $2.45 billion. The company now has 2 million paid subscribers and generates 7 million tracks per day.

Novo Nordisk announced a strategic partnership with OpenAI, unveiled April 14, 2026, to integrate AI across drug discovery, manufacturing, supply chains, and commercial operations. The deal targets acceleration of new obesity and diabetes treatments.

Anthropic reported on December 9, 2025 that the Model Context Protocol (MCP) had surpassed 97 million monthly SDK downloads across Python and TypeScript. The same announcement moved MCP's governance to the Linux Foundation's newly formed Agentic AI Foundation, co-founded by Anthropic, Block, and OpenAI. Major AI providers and coding tools have adopted MCP as a common standard.

Google's Gemini 3.1 Pro, released February 19, 2026, has posted strong results on independent evaluations, including 80.6% on SWE-bench, 94.3% on GPQA Diamond (the highest score of any model), and 77.1% on ARC-AGI-2. The model supports a context window of roughly 1 million input tokens across text, image, audio and video.

Mistral released Small 4 on March 16, 2026, a 119-billion-parameter open-source model (a mixture-of-experts model with 6 billion active parameters) with a 256k context window, configurable reasoning, and multimodal (vision) capabilities. It combines fast instruction-following, deep reasoning, and image understanding in a single unified model under an open license.

Microsoft and OpenAI rewrote their foundational 2019 deal: Microsoft ends its Azure revenue share to OpenAI, OpenAI can now sell via Amazon Web Services and Google Cloud, and the "AGI clause" is removed. Anthropic's annualized revenue surpassed OpenAI's for the first time, reaching $30B by early April 2026.

Google announced an investment of up to $40 billion in Anthropic, starting with $10 billion at a $350 billion valuation, with up to $30 billion more tied to performance milestones. Google Cloud also committed 5 gigawatts of compute capacity over five years, making it one of the largest AI investments to date.

OpenAI launched GPT-5.5, its most capable model to date, with dramatic improvements in agentic coding via Codex, deep research, and the ability to operate software end-to-end. It rolled out to Plus, Pro, Business and Enterprise users in ChatGPT on April 23, with API access following on April 24.

Anthropic released Claude Design, its first dedicated design product, as part of Anthropic Labs. Built on Claude Opus 4.7, it generates slide decks, mockups and marketing materials from conversation — with automatic brand system generation from existing codebases or design files. Available at no extra cost for Claude Pro, Max, Team and Enterprise subscribers.

Adobe unveiled Firefly AI Assistant, powered by Adobe's creative agent, on April 15, 2026. Running inside the Firefly app, it coordinates workflows across Photoshop, Premiere, Lightroom, Illustrator, Express, and Firefly from one prompt, with a public beta expected in the coming weeks. The same release added audio enhancements, including Enhance Speech, to the Firefly Video Editor, building on sound-effect generation introduced in 2025. Firefly's separately available Custom Models beta lets subscribers fine-tune generation on their own brand assets.

xAI rolled out Grok 4.3 beta to SuperGrok Heavy subscribers on April 17, 2026, announced via a post from Elon Musk on X. The model runs on the same 0.5 trillion-parameter scale as its predecessor, Grok 4.20; a 1 trillion-parameter model remains in training for a future Grok 4.4 release. Grok 4.3 reached general availability on May 2, 2026, with API pricing of $1.25/$2.50.

Anthropic rolled out Memory, a managed service for Claude Managed Agents that extracts key information from agent conversations and surfaces it when relevant — keeping context windows lean. Anthropic also announced a $100K developer hackathon on Claude Opus 4.7.

Anthropic's Claude Opus 4.7 has set a new record on SWE-bench Verified, scoring 87.6% — the highest ever for a publicly available model. Priced at $5/$25 per million tokens, it marks a step-change in agentic software engineering capabilities.

Apple's new Siri AI runs on Apple Foundation Models (AFM 3), a family of five models custom-built in collaboration with Google and based on Gemini models and cloud technology. The assistant runs on-device and via Apple's Private Cloud Compute, with the heaviest workloads processed on Google Cloud infrastructure under Apple's privacy protections.

OpenAI released GPT-5.4 on March 5, 2026, in three variants: GPT-5.4, Thinking and Pro. All three share a 1.05 million token context window — the largest OpenAI has ever released commercially. The model dynamically retrieves tool specs rather than loading every definition into the prompt.

Stanford's annual AI Index reveals that as of March 2026, Anthropic leads the Arena (LMArena) Elo leaderboard, trailed closely by xAI, Google and OpenAI. The report finds people are adopting generative AI faster than they adopted the personal computer or the internet.

OpenAI announced on April 2, 2026 that it has acquired TBPN, the Technology Business Programming Network — a daily live tech and business show that has become a cult phenomenon in Silicon Valley. It is the company's first acquisition of a media company.

PwC's 2026 AI Performance study finds a sharp divide: a small group of companies focused on growth — not just productivity — are pulling far ahead, with AI-driven performance 7.2 times higher than other firms.

OpenAI has surpassed $25 billion in annualized revenue and has taken early steps toward a potential initial public offering. Anthropic is close behind, approaching $19 billion in annualized revenue.

Anthropic announced Claude Mythos Preview on April 7, 2026, alongside Project Glasswing, a defensive cybersecurity initiative. The model will not be released publicly; access is limited to a dozen named launch partners plus more than 40 additional organizations.

Meta released Llama 4 Scout and Maverick on April 5, 2025 — the first Llama models with Mixture-of-Experts architecture. Scout features 17B active parameters across 16 experts (109B total) and a 10-million-token context window. Available for commercial use, subject to Meta's license terms.

Google released Gemma 4 in four variants: E2B, E4B, a 26B A4B mixture-of-experts model, and a 31B Dense model. The 31B Dense model ranks #3 globally on Arena AI among open models, making it the strongest open-weight model Google has shipped to date.

Chinese lab Zhipu AI released GLM-5.1 under the MIT license on April 7, 2026 — a 744-billion-parameter mixture-of-experts model with 40B active parameters per forward pass and a 200K context window. The fully open release puts serious pressure on closed-source competitors.