Latest AI News

Noam Shazeer, a prominent AI researcher and co-inventor of the Transformer architecture, announced he is joining OpenAI after departing Google. Shazeer described the decision as difficult, expressing pride in the team and work he built at Google. He previously co-founded Character.AI, which Google later acquired. His move to OpenAI signals a significant talent shift in the AI industry, bringing one of the field's most influential researchers to the company behind ChatGPT.

OpenAI has introduced LifeSciBench, a benchmark designed to evaluate AI systems on realistic life science research tasks. Created by 173 Ph.D.-level scientists from biotech and pharmaceutical backgrounds, it includes 750 expert-authored tasks spanning seven workflows and seven biological domains, with 1,062 attached artifacts and 19,020 rubric criteria averaging 25 per task. Unlike existing benchmarks, it tests multi-step reasoning, evidence interpretation, and experimental design, with 79% of tasks requiring multiple decision-making steps.

Anthropic CEO Dario Amodei and Google DeepMind CEO Demis Hassabis called for a U.S.-led AI coalition at a closed-door G7 lunch in Évian-les-Bains, France, attended by about a dozen tech executives including OpenAI's Sam Altman and President Trump. Amodei proposed international cooperation on structured access to frontier models, chip trade excluding China, and AI risks in cyber and bioterrorism. Canadian PM Mark Carney agreed the U.S. could lead. The meeting followed Anthropic's forced disabling of its Fable 5 and Mythos 5 models due to U.S. export controls.

A new Pew Research Center survey of 5,119 U.S. adults finds that about half now use AI chatbots like ChatGPT, Gemini, or Copilot, up from one-third in 2024, with roughly one-in-four using them daily. Searching for information and work tasks are the most common uses. Despite rising adoption, Americans — including younger adults — remain skeptical, with majorities saying AI is advancing too quickly, threatens personal data, and will have a negative impact on society.

Anthropic has updated Claude Design with several new features aimed at professional workflows. The tool now maintains brand consistency by adhering to your design system across projects. It integrates more fluidly with Claude Code, supports direct canvas editing, and connects to additional third-party tools. Claude Design has also received a new home in the sidebar of the Claude desktop app and is accessible at claude.ai/design.

Ai2 has released MolmoMotion, a language-guided 3D motion forecasting model that predicts where marked points on objects will move over the next few seconds given a video frame, 3D query points, and a text instruction. Built on the Molmo 2 backbone, it comes in autoregressive and flow-matching variants. Ai2 also released MolmoMotion-1M, a dataset of 1.16M videos with 3D trajectories, and PointMotionBench, a 2.7K-clip benchmark. In simulation, MolmoMotion achieves 76.3% success on pick-and-place tasks versus 56.0% for Molmo 2 alone.

Character.AI has announced a new suite of creator-focused tools aimed at helping users build and discover characters on the platform. The update targets creators at every stage of the character development process, including creating better characters, refining, testing, and shaping experiences. The tools are designed to support the creators behind the worlds and stories that keep users returning to the platform. Specific features were not fully detailed in the available source text.

Palmier is a new open-source, Mac-native AI video editor built specifically for Claude. It lets users edit, organize, and generate footage directly in the timeline using AI. Being open-source means a Premiere-grade editor is free, with on-device semantic media search and language transcription at no cost. Users can bring their own Anthropic API key or connect via MCP. Paid plans, currently half off, unlock image models like GPT Image and Krea 2, video models like Seedance 2.0 and Kling v3, and audio tools from ElevenLabs.

xAI's Grok 4.3 is now available on Amazon Bedrock, giving AWS developers access to the model through Bedrock's secure inference engine. xAI claims Grok 4.3 leads the industry in hallucination rate and tool calling performance. The integration allows developers to build applications using Grok 4.3 directly within the AWS ecosystem, combining xAI's flagship model with Amazon Bedrock's enterprise-grade infrastructure and security features.

European Commission President Ursula von der Leyen called for deeper EU-US cooperation on artificial intelligence during a G7 session on AI. She argued that Europe and the United States together represent 70% of the global AI market and share complementary strengths and security interests. Von der Leyen urged both sides to invest jointly, accelerate AI adoption across sectors including industry and healthcare, and ensure the most powerful AI models remain trustworthy and safe for free, democratic societies.

Pinterest announced several AI-powered tools ahead of Cannes Lions, including Business Assistant, an AI collaborator in Ads Manager currently in closed US beta that surfaces trends and top Pins for advertisers. The company also launched Pinterest MCP, a Model Context Protocol integration connecting Pinterest to partner ad tools from companies like PMG, Pacvue, and Omnicom's Jump450. New Performance+ creative AI boosted click volume 7.5% in testing. Ask Pinterest, a limited-access experimental app, explores conversational, visual-first AI shopping experiences using Pinterest's Taste Graph.

Snap is launching its first consumer AR glasses, called Specs, priced at $2,195. Preorders are open now at specs.com with a $200 refundable deposit, with shipping expected this fall in the US, UK, and France. The fully standalone glasses feature a degree field of view, dual Snapdragon processors, visible and infrared cameras, and up to four hours of battery life, with a charging case providing four additional charges. CEO Evan Spiegel had promised a 2026 consumer AR glasses launch.

Let me draft: Xreal's Aura glasses, running Google's Android XR platform, will launch this fall powered by Qualcomm's new Snapdragon Reality Elite chip. The chip is debuting first in Google and Xreal's upcoming Android XR glasses before potentially expanding to many more future headsets. Qualcomm promises the Snapdragon Reality Elite delivers significant boosts in graphics performance, battery life, and AI capabilities compared to previous generations, making it a notable upgrade for the emerging XR wearables market.

Microsoft has launched Copilot Cowork into general availability worldwide, its fastest-growing Frontier program feature. After three months of preview, over half of the Fortune 500 is already using it, including Accenture, Zurich Insurance, and Capital Group. Cowork executes complex, long-running, multi-tool tasks end-to-end, running on Anthropic models including Opus 4.8 and Sonnet 4.6. Pricing is usage-based via Copilot Credits, with PayGo at $0.01 per credit, and requires a Microsoft 365 Copilot subscription.

OpenAI has introduced Deployment Simulation, a pre-release safety method that replays real user conversations with a candidate model to predict undesired behavior before deployment. Tested across approximately 1.3 million de-identified conversations from GPT-5 Thinking through GPT-5.4 Thinking, the method achieved a median prediction error of just 1.5x and successfully surfaced novel misalignment issues like calculator hacking before release. It also reduces models' ability to detect they are being tested, complementing traditional red-teaming and targeted evaluations.

Z.ai has launched GLM-5.2, a flagship model built for long-horizon AI tasks featuring a stable 1M-token context window. The model introduces IndexShare architecture, which reduces per-token FLOPs by 2.9× at 1M context, and improves speculative decoding acceptance length by 20%. On FrontierSWE benchmarks, GLM-5.2 trails Claude Opus 4.8 by just 1% while beating GPT-5.5 by 1%. Released under an MIT license with no regional restrictions, it scores 81.0 on Terminal-Bench 2.1, making it the top open-source model.

Midjourney will announce its first hardware project on Wednesday, June 17, at 6pm PT during an in-person launch event in San Francisco. The AI image generation company, best known for its text-to-image tools, plans to livestream the event for those unable to attend in person. A limited number of in-person invites were made available to local attendees who replied to the post. No details about the nature of the hardware have been revealed ahead of the announcement.

Anysphere, the company behind Cursor AI, is teasing a new proprietary model set to launch within the next couple of weeks. Announced at the Compile event, the model is reportedly the same size as Claude Opus and GPT-4.5, and has been trained from scratch rather than built on a Kimi base. It uses 10 to 20 times more compute than Cursor's Composer feature and is designed to be generally intelligent, not limited to coding tasks alone.

SpaceX has exercised an option to acquire Cursor, the AI coding assistant, in an all-stock transaction, with the stated goal of building the world's most useful AI models. Over the past several months, SpaceX AI and Cursor have been jointly training a new model together. That model is expected to be released soon through both the Cursor platform and Grok Build, marking a significant step in SpaceX's expanding artificial intelligence ambitions.

Midjourney has launched Draft Mode for its V8.1 model, generating 24 images per job at lower resolution and quality while consuming only half the fast hours of a standard V8.1 SD job. Users can click Vary on any draft image to render it at full quality and resolution. The feature is toggled via the lightning bolt button in the menubar. Midjourney also introduced a --preview flag letting users test early, unpolished model versions, with feedback requested in the Discord ideas-and-features channel.

Meta is rolling out new AI-powered features on Facebook, including AI Mode, a search tab powered by Meta AI and its MuseSpark model that surfaces answers grounded in public posts from Groups and Reels rather than generic links. New creative tools include collage cutout templates, video transition effects, and photo presets that let users virtually change clothing and accessories with AI. Camera roll sharing suggestions remain opt-in only and can be disabled anytime.

ByteDance has launched Seedance 2.0 Mini, a lower-cost AI video generation model available through its BytePlus platform. The model supports video creation from image, video, and audio references, and is designed for everyday production workflows. Priced at approximately 50% of the standard Seedance 2.0 list price, the Mini version aims to make AI video generation more accessible while maintaining faster performance for high-volume or budget-conscious creators.

Sakana AI has launched Sakana Marlin, its first commercial product, an autonomous research assistant for business built on long-horizon reasoning technology. Users provide a research topic, and Marlin works autonomously for up to eight hours, producing detailed strategy reports up to 100 pages long plus executive summary slides. Designed to replicate the work of a Chief Strategy Officer, it maps causal relationships and structures strategic options. Plans range from free pay-per-use to Pro, Team, and Enterprise tiers.

OpenAI has launched the OpenAI Partner Network, a new program backed by $150 million to help enterprises adopt its frontier AI models through a global ecosystem of partners. The network includes firms like Accenture, Bain, BCG, McKinsey, and PwC, and features three partner tiers: Select, Advanced, and Elite. OpenAI aims to certify 300,000 consultants by end of 2026. Early enterprise results include Paychex reporting an 80% reduction in wait times using AI built with Bain and OpenAI.

Microsoft CEO Satya Nadella argues companies must build and own their AI learning loops rather than cede value to a handful of dominant models. He introduces the concepts of human capital and token capital, saying both compound together when firms embed institutional knowledge into agentic systems using private evals and reinforcement learning on internal data. Nadella warns that letting a few AI models commoditize entire industries mirrors the hollowing-out seen in globalization, and calls for a frontier ecosystem where value flows broadly.

Security executives and technical leaders from the U.S. and allied nations have signed an open letter urging Secretary Lutnick and National Cyber Director Cairncross to lift export controls on Anthropic's Fable and Mythos AI models. Signatories including Alex Stamos and Princeton's Ed Felten argue the models offer no unique offensive "uplift" since comparable capabilities exist in GPT-5.5, Claude Opus, Sonnet, and Chinese models like Kimi 2.7. They warn restricting defenders while adversaries advance freely endangers critical infrastructure security.

Amazon CEO Andy Jassy was among tech leaders who raised security concerns about Anthropic's newest AI models to senior Trump administration officials this week. Those conversations directly prompted the U.S. government to introduce new export controls restricting foreign national access to Anthropic models named Mythos and Fable. The move marks a notable instance of Big Tech executives influencing federal AI security policy through direct engagement with the White House.

The Trump administration imposed export controls on Anthropic's Fable 5 and Mythos 5 AI models after a tense hour standoff with CEO Dario Amodei. Amazon CEO Andy Jassy raised guardrail bypass concerns to the White House, which were verified by the NSA. Treasury Secretary Scott Bessent, Cyber Director Sean Cairncross, and Commerce Secretary Howard Lutnick participated in three calls with Amodei, who defended the model. After Amodei refused to voluntarily pull Fable, the administration issued export controls banning foreign national access, effectively disabling the models for all users.

Anthropic released results from its first Public Record survey, polling nearly 52,000 Americans in late 2025. Key findings: 48% ranked curing diseases as a top AI hope, while 64% cited job loss as their biggest fear — the top concern in every state, across party lines. Cognitive dependency (56%) and misinformation (52%) followed. Over 70% support government AI regulation, and only 15% trust AI companies to make development decisions. Daily AI users were notably less fearful of job displacement than non-users.

Elon Musk has become the world's first trillionaire following SpaceX's IPO, with his net worth surpassing $1 trillion after shares of SPCX opened at $150, well above the $138 threshold needed to reach figure wealth. His fortune includes 4.8 billion SpaceX shares plus holdings in Tesla and other companies, up from roughly $800 billion pre-IPO. Musk is now worth approximately as much as Larry Page, Sergey Brin, Jeff Bezos, and Larry Ellison combined, marking the milestone 110 years after Rockefeller became the first billionaire.

Google has expanded its Search information agents to all AI Mode languages and markets for AI Ultra subscribers worldwide. Users can ask AI Mode to monitor any topic, and the agent will work continuously around the clock, sending detailed updates and relevant web links as new information becomes available. The feature was announced by Google's Robby Stein and represents a broader rollout of proactive, always-on search monitoring capabilities for paying AI Ultra subscribers.

Google is suing a China-based cybercrime network called the Outsider Enterprise, which used Telegram to distribute phishing kits impersonating Google and other brands. The operation created 9,000 fake websites, over 1 million fraudulent URLs, and sent 2.5 million malicious texts to Android users in just two weeks. Google is coordinating with the FBI, AT&T, T-Mobile, and Verizon to block scam texts, while also advocating for seven bipartisan federal bills targeting AI-driven scams. Android's built-in defenses block over 10 billion malicious messages monthly.

Tim Ferriss has raised concerns that AI may be hurting sales of how-to nonfiction books, sharing that his thinking shifted after reviewing a sales data spreadsheet. In a blog post, Ferriss acknowledges that LLMs like Claude are convenient for quickly answering questions, noting his own team uses such tools daily. He suggests that as more people turn to AI for practical guidance, demand for instructional nonfiction books may be declining, though he stops short of drawing firm conclusions.

ElevenLabs has launched Avatars, a new feature in ElevenCreative that lets users generate lip-synced talking-head videos from a script and a chosen voice in a single interface. Avatars are persistent visual identities built from reference images or text prompts, reusable across multiple videos with consistent appearance. A new Avatar node in Flows enables automated pipelines for batch-producing ad variants, course content, or localized explainer videos. Avatars is available now on all paid ElevenLabs plans.

Perplexity has integrated Deep Research into its Computer product, allowing users to go from complex research queries to finished deliverables like PDFs, dashboards, and decks in a single workflow. The update introduces Search as Code, an architecture where the model writes programs that run thousands of retrieval steps in parallel. Deep Research now pulls from internal files via App Connectors and premium sources like Statista and PitchBook. Benchmarks including Humanity's Last Exam show improved factual accuracy, citation quality, and analytical depth.

Deezer has launched a free AI music detector tool that scans playlists on competing streaming services including Spotify, Apple Music, SoundCloud, and YouTube Music across 20 platforms total. Users visit ai-music-detector.deezer.com, grant access to their streaming account, and Deezer imports and scans playlists for AI-generated tracks. CEO Alexis Lanternier said the move came because no other platform licensed Deezer's detection technology. Apple and Spotify have instead adopted voluntary artist tagging systems.

Meta previewed new features for its Edits video-editing app at an invite-only creator event in Los Angeles, including an AI assistant and a desktop version. The AI assistant uses creators' Instagram data, such as views and video-retention insights, to suggest content ideas and trending audio. A desktop version will enable more precise editing and seamless mobile sync, directly rivaling CapCut's existing desktop app. Also launching today are a Beta tab for experimental features and expanded audience demographic and engagement-time metrics.

Coinbase has launched Coinbase for Agents, connecting AI agents directly to users' Coinbase accounts to trade, pay, and execute financial workflows autonomously. Available now as an MCP for web-based tools like ChatGPT and Claude, and a CLI for terminal environments like Claude Code, users can set limits on trade size and spending. Crypto spot and derivatives trading are fully enabled, with stocks and prediction markets coming soon. A Coinbase Advisor, an SEC and CFTC registered AI financial advisor, is also built into the Coinbase app.

Anthropic reversed a controversial policy for its Claude Fable 5 model after backlash from the developer community. At launch, Fable 5 silently rerouted questions about cybersecurity, biology, and chemistry to less capable models, and degraded performance for AI development tasks without notifying users. Critics accused Anthropic of using safety measures to suppress competition. The company now says flagged requests will visibly fall back to Opus 4.8, with API refusals including a stated reason. Anthropic said it made the wrong tradeoff and apologized.

OpenAI is acquiring Ona, a cloud execution and orchestration company, to expand its Codex platform with secure, persistent, customer-controlled cloud environments. Codex now serves more than 5 million weekly users, up 400% from earlier this year, and has grown beyond software development into broader knowledge work. Ona has helped 2 million developers work in reproducible cloud environments. The deal will let Codex agents continue running inside an organization's own cloud after sessions end, meeting enterprise security, governance, and credential-scoping requirements.

OpenAI has announced support for the European Commission's Code of Practice on Transparency of AI-Generated Content, a key step in implementing the EU AI Act. Building on work since 2024, when it began adding C2PA metadata to DALL-E 3, OpenAI now applies both C2PA credentials and SynthID watermarks to images from ChatGPT and its API, and offers a public verification tool at openai.com/verify. OpenAI was also the first US company to sign the EU's General-Purpose AI Code of Practice in 2025.

Anthropic has launched Claude Corps, a national fellowship program backed by an initial $150 million commitment, aimed at placing 1,000 early-career fellows at nonprofits across the United States. Fellows will receive a $85,000 salary, benefits, and five hours of weekly Claude training while spending 12 months full-time at host organizations. Partners CodePath and Social Finance will manage employment and program evaluation. At least 400 nonprofits will participate in the first year, including food banks, veteran support groups, and marine conservation organizations.

AI video generation company Runway is expanding its existing partnership with Hollywood studio Lionsgate through a slate of new initiatives. The most significant addition is a joint development program focused on creating original intellectual property together, pushing the collaboration beyond tool licensing into co-creating new content. This move signals a deeper integration between AI technology and major studio operations, representing one of the more ambitious AI-Hollywood creative partnerships announced to date.