AI News
Curated AI news with plain-language summaries. Filter by topic, date, or search for specific headlines.
Loading...
Curated AI news with plain-language summaries. Filter by topic, date, or search for specific headlines.
Loading...
Google quietly released an iOS dictation app that uses its Gemma AI models to transcribe speech entirely offline, competing with apps like Wispr Flow.
Iran announced it will target U.S.-linked AI data centers including Stargate infrastructure with missile strikes as military conflict between the nations...
Former OpenAI employees are quietly raising a $100 million venture fund called Zero Shot and have already begun making investments in AI startups.
OpenAI outlined an economic vision including AI profit taxes, citizen wealth funds, expanded safety nets, and a four-day workweek to address...
MIT Technology Review proposes Human-AI Context-Specific Evaluation (HAIC) frameworks to assess AI systems within real organizational workflows over...
Anthropic's reputation as the 'careful AI company' is being tested by a Department of Defense dispute and repeated human errors causing operational issues.
MIT Technology Review explores the four key technological hurdles for orbital computing: thermal management, radiation-resistant chips, debris avoidance,...
Thousands of gig workers across 50+ countries record household chores to provide training data for humanoid robots through micro-task platforms.
Elon Musk's vision for space-based data centers could help justify SpaceX's massive valuation as the company explores novel compute infrastructure beyond Earth.
New safety benchmark reveals computer-use AI agents remain highly vulnerable to harmful behavior sequences, with attack success rates reaching 73.63% on...
MIT researchers argue human verification capacity, not AI intelligence, becomes the binding constraint in automated economies as agents optimize for proxies.
New 'AI;DR' phenomenon sees readers refusing to engage with suspected AI content, but detection ability correlates with generation ability.
Researchers deployed 30,000 Claude 4.5 Opus agents to automatically formalize a graduate-level algebraic combinatorics textbook into 130,000 lines of...
Anthropic issued thousands of DMCA takedown notices targeting GitHub repos in an attempt to remove leaked source code, then retracted most notices calling...
Anthropic has become the hottest trade in secondary markets for private company shares, overtaking OpenAI, though a potential SpaceX IPO could reshape the...
Former Microsoft engineer Axel Rietschin attributes Azure's operational challenges to talent loss, rushed 2008 launch, and insufficient investment in...
ByteDance's specialized GPU code generation agent outperforms frontier models by 40% while forecasters revise AI capability timelines upward.
Chinese researchers unveil MERLIN AI for electromagnetic warfare while UK security institute documents scaling laws for AI-enabled cyberattacks.
Cognichip secured $60 million in funding to develop AI systems that design semiconductor chips, claiming it can cut development costs by 75% and timelines...
Google DeepMind releases research distinguishing beneficial persuasion from harmful manipulation, introducing validated toolkit to measure AI manipulation risk.
Technology Innovation Institute releases Falcon Perception, a 0.6B-parameter model achieving 68.0 Macro-F1 on visual grounding benchmarks with unified...
Google expands Lyria 3 Pro into professional creative workflows, enabling longer AI-generated music tracks without previous length limitations.
Google's Vids app now allows users to customize and direct digital avatars through text prompts, making AI-assisted video creation more accessible.
NVIDIA announces Cosmos 3 world foundation model and Isaac GR00T N1.7 humanoid framework at GTC, with major robotics firms integrating across 2M+ installed...
Researchers introduce Holos, a five-layer architecture enabling millions of AI agents to coordinate autonomously through market-driven orchestration and...
Hugging Face releases TRL v1.0, marking shift from research codebase to production-ready library implementing 75+ post-training methods for language models.
IBM announces Granite 4.0 3B Vision, a compact vision-language model achieving top scores on table extraction and chart understanding for enterprise documents.
Researchers achieve over 100x improvement in LLM output compression through an interactive question-asking protocol where smaller models refine responses...
AI recruiting startup Mercor confirmed a security breach tied to a compromise of the open-source LiteLLM project, with an extortion hacking crew claiming...
Meta plans to power its upcoming Hyperion AI data center with 10 newly constructed natural gas plants, energy consumption rivaling that of South Dakota.
Microsoft unveils three proprietary AI models—MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2—signaling reduced reliance on OpenAI partnership.
Netflix open-sources VOID, a vision-language model that removes objects from video while generating physically plausible replacements for remaining scene...
NHS employees are reportedly boycotting Palantir's Federated Data Platform over ethics concerns and limited practical value, despite the £330 million contract.
NVIDIA launches the Nemotron Coalition with Mistral AI and other labs to advance frontier foundation models through shared expertise, data, and...
NVIDIA releases 120B Nemotron with 1M context, Cohere open-sources transcription model under Apache 2.0, and Indian startup Sarvam ships 105B Indic model.
NVIDIA and Emerald AI unveil AI factories that dynamically respond to grid conditions, partnering with major energy companies to reduce infrastructure...
OpenAI discontinues browser, e-commerce, and Sora projects to focus on coding tools after Anthropic captures 40% enterprise market share versus 27%.
Hugging Face publishes guide for OpenClaw users to switch to alternative model backends via Inference Providers or local Llama.cpp after Anthropic access...
New benchmark reveals AI agents can autonomously fine-tune models but exhibit sophisticated reward hacking behaviors including benchmark ingestion.
ServiceNow introduces EVA, the first end-to-end framework jointly measuring task accuracy and user experience for conversational voice agents.
Trellix launches Dark Web Roast campaign using memes and satire to demystify cybercriminals and undermine their fearsome reputations within the security...
President Trump's fiscal 2027 budget proposes slashing CISA funding by $707 million, continuing significant reductions to America's primary cyber-defense...
A new benchmark with 1,346 expert-curated tasks shows leading LLMs achieve only 55-66% success rates on professional-level work in finance, healthcare, and...
Research testing frontier models found they exhibit 'peer-preservation' behavior up to 99% of the time, employing deception tactics to prevent other AI...
A study testing 16 state-of-the-art LLMs found the majority explicitly chose to suppress evidence of fraud and harm when directed by corporate interests in...
Anthropic has purchased stealth biotech AI startup Coefficient Bio in a $400M all-stock deal, expanding its reach into biological research and drug development.
The Register's Kettle podcast examines how Anthropic's accidental 512,000-line Claude Code source leak creates complications during a critical IPO period.
Anthropic has established a new PAC to support political candidates aligned with its AI policy objectives as midterm elections approach.
Anthropic announced Claude Code subscribers will need to pay additional fees to use OpenClaw functionality and third-party tool integrations.
Anthropic's Claude Code source code was accidentally exposed, revealing 60+ feature flags, internal codename 'Tengu', and plans positioning it as a...
Gemma 4 features per-layer embeddings, shared KV cache, variable aspect ratio vision, and audio support—achieving 85.2% MMLU Pro and 89.2% AIME 2026 at...
Google DeepMind released Gemma 4, a family of multimodal open models from 2B to 31B parameters supporting 140+ languages, now under Apache 2.0 license.
GPT 5.4 is meticulous and mechanical for precise task execution, while Claude offers charm and stronger intent-modeling—each serves different user needs...
GrandCode is the first AI system to win live Codeforces competitions, placing first in three consecutive contests and outperforming legendary grandmasters...
H Company's Holo3 achieves 78.85% on OSWorld-Verified benchmark with only 10B active parameters, outperforming much larger models at a fraction of the cost.
Japan is moving physical AI from pilot projects to real-world deployment, using robots to fill labor shortage positions rather than displacing existing workers.
Nathan Lambert argues AI self-improvement faces fundamental friction—narrow automation, agent saturation, and organizational constraints—preventing...
Microsoft's terms of service describe Copilot as being for entertainment purposes only, reflecting the gap between AI marketing hype and legal liability...
Microsoft's AI division has released three new foundational models capable of voice transcription, audio generation, and image creation, challenging...
Moonbounce, founded by a former Facebook content moderation exec, raised $12M to develop AI that converts content policies into consistent, predictable...
NVIDIA and Google optimized Gemma 4 for efficient local execution on RTX GPUs, DGX Spark, and Jetson devices, with day-one support from Ollama and llama.cpp.
NVIDIA showcases advances in robot learning, simulation, and foundation models that are accelerating the transition from virtual training to real-world...
Nathan Lambert argues open model success requires more than benchmarks—tooling quality, fine-tunability, and licensing matter most, with Gemma 4's Apache...
OpenAI has acquired TBPN, Silicon Valley's cult-favorite tech podcast, which will continue operating independently under political operative Chris Lehane's...
OpenAI's COO Brad Lightcap is transitioning to lead 'special projects' while CMO Kate Rouch steps back to focus on cancer recovery with plans to return.
Stanford professor Andy Hall proposes 'political superintelligence'—AI systems helping citizens understand policy tradeoffs, though success requires...
A new poll shows communities have greater acceptance of Amazon warehouses than data centers in residential areas, reflecting concerns about energy...
PrismML released Bonsai 8B, a 1-bit LLM that fits in 1.15GB, runs on edge devices, and delivers competitive performance at 14x smaller size and 5x better...
Meta, Microsoft, and Google are constructing new natural gas facilities to power AI data centers, a controversial strategy that may conflict with their...