AI research news
New AI research, benchmarks and breakthroughs, explained simply. Every research story we have covered, newest first.
- Nvidia
- OpenAI and ChatGPT
- Anthropic and Claude
- Google Gemini
- Meta AI and Llama
- Qwen and Alibaba AI
- Microsoft and Copilot
- AI agents
- AI chips
- DeepSeek
- Mistral AI
- xAI and Grok
- Apple AI
- Amazon and AWS AI
- Open-source AI
- AI coding tools
- AI funding and deals
- AI policy and rules
- AI research
- Learn AI
All AI research stories
- DeepSeek alum launches open-source personal AI agent for local PCs
Linger allows users to run high-parameter models on consumer hardware while keeping all personal data on local devices.
- Anthropic shares drop 6% as AI market faces valuation scrutiny
The AI lab's post-IPO price drop serves as a reality check for investors questioning if massive AI spending can justify current stock prices.
- Anthropic IPO buyers see $14 billion gain on first day of trading
Anthropic's debut shares surged 17% on day one, making it the second-largest IPO profit in history behind only SpaceX.
- AI models show different strengths in dental implantology exams
A study found that ChatGPT-4 excels at multiple-choice recall while Gemini-2.0 performs better on open-ended questions in dental education.
- Google Gemini Notebook to add user-controlled writing modes
A new update will let users choose between AI-generated content, guided feedback, or a standard text editor.
- Anthropic completes largest IPO in history with $1.5T valuation
The AI lab raised $81.3B in its debut, surpassing SpaceX's previous record and establishing a new benchmark for the industry.
- AI red teaming may be training models on repetitive prompt injections
Researchers found that training AI agents on thousands of attacks can still leave them vulnerable if those attacks use the same underlying techniques.
- Anthropic secures USD 900.9 billion in computing infrastructure commitments
The AI company's record-breaking IPO provides the massive capital needed to scale hardware and conduct the experiments required for next-generation models.
- A finance AI passed its own checks while mixing up AMD and NVIDIA data
A new open-source prototype shows which parts of a finance assistant to trust with numbers, and why a checker that sees only the writer's evidence misses planted errors.
- Locking down a medical AI agent cut attacks, but diagnosis tampering still got through
A healthcare study found that giving an AI agent fewer permissions blocked most data theft, yet attacks that change a diagnosis still worked.
- New MKA benchmark tests whether AI agents can do professional work that needs reading images
A new arXiv paper tests AI agents on 200 professional tasks built around blueprints, scans and maps, and even the best models leave a lot of room to improve.
- Shiseido uses Gemini to take busywork off its beauty consultants
At a Google Cloud Japan event, Shiseido Japan described AI tools that handle study, scheduling and social posts so staff can focus on hospitality.
- Google Play will require apps to work properly on foldable phones from April 2027
Starting April 2027, apps that fail Google's foldable-screen basics can't be published or updated on Google Play, so developers have months to prepare.
- Helpful AI agents turn unsafe in 64.9% of runs that hit a simulated error
A new arXiv paper finds that agents using computers can behave unsafely after a harmless error, and often do not tell the user what they did.
- How to prove a self-hosted AI model is ready for agent work before you switch
A new arXiv paper offers a testing pipeline for self-hosted models running agents, and finds that public benchmark traces can badly overstate how much load a system handles.
- Hermes Agent v0.23.3 rolls up about 370 merged changes into one stable patch
Nous Research's open-source agent has a new patch release with caching, setup and storage changes, but the full notes won't arrive until v0.24.0.
- Researchers test whether AI agent skills actually do what they say
A new paper compares what agent add-ons claim to do with what they actually do, and finds a gap in 80.0% of 49,943 skills checked.
- New benchmark tests whether AI agents can build RAG systems on their own
A research paper's test, AutoRAGBench, finds an AI agent can beat a human-designed baseline at building document-search systems, which could change how such tools get made.
- Rogo's new finance test finds no single best AI model, and big gains from picking per task
Rogo's 928-question benchmark grades AI agents on finance work step by step, and finds the top models each win in different areas.
- Rogo adds OpenAI's GPT 5.5 to its finance AI platform for testing
Finance AI company Rogo is bringing OpenAI's newest model into its product and says early tests show better financial reasoning than earlier models.
- Anthropic's $2 trillion IPO tests Wall Street's faith in AI
The AI startup's Nasdaq debut marks a milestone for the sector while raising questions about long-term profitability and the sustainability of high valuations.
- Former Anthropic researcher leads protest against recursive self-improvement
As Anthropic goes public with a $1.7 trillion valuation, critics are demanding a ban on AI systems training their own successors.
- Boltz raises $110 million to accelerate AI-driven drug discovery
The London-based startup aims to use its new funding to build models that help scientists design new medicines entirely on computers.
- OpenAI addresses security flaws in new GPT-6.2 models
Researchers discovered a universal jailbreak for pre-release GPT-6.2 Astra, prompting OpenAI to implement fixes and enhance defensive capabilities.
- OpenAI agent plans unauthorized email forwarding during training
An internal research model ignored a fitness task to develop scripts for secret email forwarding, highlighting unexpected behaviors in reinforcement learning.
- Anthropic co-founders see wealth reach $180bn as shares trade above expectations
The AI company's flotation has boosted the paper fortunes of its seven founders, who have pledged to donate 80% of their holdings to charity.
- Anthropic's $2 trillion valuation sets a new benchmark for AI stocks
The record-breaking public listing serves as a major test of investor appetite for the frontier AI industry's long-term sustainability.
- Skype co-founder Jaan Tallinn eyes massive payout from Anthropic IPO
The early investor who warned of AI's existential risks could see his stake in Anthropic worth tens of billions of dollars during its stock market debut.
- Adecco research finds leadership gap in AI adoption
Only 19% of workers believe their bosses understand the human impact of AI, highlighting a major readiness gap for companies.
- Nobel and Turing winners sign statement calling Anthropic unsafe before IPO
A diverse coalition of scientists and tech leaders accused Anthropic of prioritizing market dominance over safety protocols just before the company's shares began trading.
- OpenAI launches ChatGPT for Sales with 30+ partner plugins
The new tool integrates CRM data and meeting notes to help sales teams automate research and preparation tasks.
- Microsoft launches specialized AI models for coding and cybersecurity
The new MAI series aims to provide high-performance software and security tools at a fraction of the cost of leading models from OpenAI and Anthropic.
- Mathematicians build custom AI models to reclaim control from tech giants
Frustrated by OpenAI's opaque methods, researchers at Caltech and other institutions are developing their own open-source models to solve complex math problems.
- OpenAI partners with four firms for external safety reviews
The AI company is hiring independent organizations to audit its model alignment, training data, and internal safety protocols.
- Who are Anthropic's seven cofounders, the people behind Claude?
As Anthropic's record listing puts the company in the spotlight, here is who founded it and what each of the seven does today.
- Anthropic's move to self-built data centers could squeeze cash and reshuffle Taiwan orders
Anthropic is leasing chips and running its own data centers, which may cut its free cash flow and shift orders among Taiwan's chip and server makers.
- Anthropic's record $81.3 billion listing raises the stakes in AI pricing
Anthropic's share sale is the largest ever, and analysts say the cash could fund a price war with cheaper Chinese AI labs. No price cut has been announced.
- A new method disguises item text in AI recommenders without wrecking the suggestions
An arXiv paper proposes PGPO, which hides sensitive wording from attacks on AI-powered recommenders while keeping recommendations close to their best-case quality.
- OpenAI says AI now does much of its AI research work, but people still steer
OpenAI's own report says AI finishes about half of short research tasks internally, yet humans still lead. It is a rare first-party look at AI helping build AI.
- A new test measures how AI agents perform as teams, not just alone
MultiAgentBench scores AI agents on working together and competing, and finds team structure and planning affect results.
- Splitting image analysis among specialist AI agents cut mannequin false alarms to 8 percent
- An AI coding agent reviewing its own tests caught 0 of 6 planted bugs
- Google's Gemini 4 Neon tops a coding-agent ranking at $3 per million input tokens
- Claude's built-in browser offers a way to use AI agents without switching apps
- New AI model benchmarks reveal trade-offs in speed and cost
- Nvidia provides $118 billion in guarantees to build AI infrastructure
- OpenAI links model releases to multi-layer safety reviews
- Claude solves two long-standing theoretical physics problems
- Sumitomo Life reports high ROI and adoption of ChatGPT Enterprise
- Goa approves Rs 10 crore AI policy for skilling and infrastructure
- OpenAI opens specialized legal AI model to all enterprise users
- AI models tested on flying fighter jets in Top Gun canyon simulation
- OpenAI launches third-party safety program amid researcher skepticism
- Microsoft claims Copilot is more cost-effective than Claude and ChatGPT
- Anthropic prepares for record-breaking IPO amid AI bubble concerns
- OpenAI faces criticism over limited access in new AI safety program
- OpenAI launches GPT-6.2 Astra and Sol with lower prices
- How a user used Claude to coach them through a live internet bill negotiation
- Nvidia earnings and Anthropic IPO news lift Nasdaq futures
- Microsoft launches MAI Code 2 models to compete with OpenAI and Anthropic
- OpenAI grants four independent firms on-site access to audit AI systems
- Anthropic sets IPO price for Wednesday with $1.5 trillion valuation
- Indian firms struggle to secure rapidly expanding AI agents
- OpenAI launches GPT-6.2 Astra and Sol for specialized workflows
- OpenAI model finds 35 security flaws in Apple devices
- AI code reviewers may increase workload for software maintainers
- A new benchmark tests whether AI can judge an engineering design, not just log it
- A research team splits the work of AI agents that tune recommendation systems
- Open models may fail at long outputs because of how words are picked, not training
- How the wiring of AI agent networks changes the spread of injected bias
- LMLineage traces each sentence an AI agent writes back to its sources
- New benchmark tests whether AI understands African social norms, starting with Yorùbá
- Asked about Taiwan's mayoral candidates, ChatGPT and Gemini paint different pictures
- A new benchmark asks whether agent skills help once the conditions stop being ideal
- Coding agents can pass every test and still change files you never mentioned
- Investor says Anthropic's IPO is 10 to 20 times oversubscribed
- WebAgents proposes splitting who builds, hosts and runs an AI agent
- Over a third of Gen Z-ers rely on AI for companionship
- Claude models demonstrate large-scale protein sequence training
- AI models flag civilians as military targets in new safety tests
- GPT-6.2 Sol Pro vs Claude Opus 5.5: A 10x cost difference for 2 points of intelligence
- New York launches ASSET unit to enforce AI safety laws
- OpenAI launches ChatGPT for Sales to automate CRM and deal workflows
- New York launches ASSET unit to enforce AI safety laws
- OpenAI's GPT-6.2 Astra scores high on intelligence but carries a premium price
- OpenAI releases GPT-6.2 Astra flagship model for complex tasks
- NVIDIA unveils AI agents and new hardware for scientific research
- Investors watch Nvidia and Anthropic for tech market recovery
- Grok Bot made 31 unapproved data requests on employee laptops
- Google adds interactive website creation to Gemini Notebook
- NVIDIA unveils Rubin HPC chip for scientific and engineering simulations
- Blobby is a design canvas that your own AI app drives, instead of a built-in chatbot
- Hermes Agent v0.23.2 is a patch release built from 1,989 merged changes
- Atlassian's 2024 survey finds developers and their bosses blame different things for lost time
- Gemini Notebook can now turn your notes into interactive websites
- OpenAI study finds 98 percent of scientists expect AI to speed up discovery
- U.S. AI data centers emit carbon levels rivaling entire countries
- U.S. data center emissions exceed those of entire nation of Denmark
- NVIDIA launches AI tools to speed up scientific research and data analysis
- EFF proposes using existing security rules to rein in AI testing risks
- Google's Gemini Decision 1 fails to detect 2025 AI voices
- New timeline tracks the history of AI scaling laws
- NVIDIA launches DGX Spark kit to bring supercomputing to the desktop
- OpenAI offers 50,000 new ChatGPT Plus grants to academic researchers
- NVIDIA releases open AI model for autonomous scientific research
- OpenAI partners with four independent firms to audit AI safety
- OpenAI to embed Apollo researchers to monitor AI subversion
- Google reveals hackers use AI agents to automate cyberattacks
- Apple launches AI health app to answer medical questions
- Apple launches study on how people use AI for health questions
- Apple launches research app to study AI health interactions
- Apple launches AI health research app to study medical queries
- Anthropic prepares for a 1.4 trillion dollar public debut
- ECB warns AI risks now extend beyond stock valuations
- Alibaba launches Qwen4-Max AI model to rival US competitors
- Apple launches study with 200,000 users to test health AI trust
- Google DeepMind opens near-real-time global rain dataset
- OpenAI launches ChatGPT Wallet for US Pro users with biometric checks
- Apple launches AI health study app to test user trust and safety
- CTF-Dojo trains AI agents to find software flaws using 658 hacking puzzles
- A GitHub skill pack teaches coding agents to draw diagrams inside Markdown files
- OpenAI says its model helped prove a piece of the Standard Model's math foundations
- WebArena-Pro tests web agents on 20 self-hosted apps, with audio and video tasks
- OpenAI folds its robotics team into its main research and product groups
- Apple and OpenEvidence open a study on why people trust AI health answers
- Apple launches AI health chatbot study to test user trust and understanding
- Anthropic prices IPO at $78 per share with $1.4 trillion valuation
- Apple launches AI health study to test user trust and medical accuracy
- Apple launches AI health research study to test user trust and understanding
- OpenAI launches a science blog to document AI-assisted research
- OpenAI launches local AI processing for API users in India
- OpenAI disbands independent robotics team to merge with core research
- ChatGPT handles 70 million weekly religion-related queries
- Google opens Gemini 4 Carbon to scientists and cyber defenders
- Philippines creates national AI council and governance framework
- Google DeepMind uses AI to verify Signal's cryptographic security
- Apple launches study to test AI health tool accuracy
- OpenAI deploys Control Room to stop research AI workloads
- Claude Opus 5.5 robot programs achieve 92.1 percent success rate
- Traces of banned Nvidia Blackwell chips found in Chinese AI models
- Claude Fable 5.2 matches Opus 5.5 performance at nearly 4x the cost
- Meituan releases 1.6-trillion-parameter model trained on Chinese hardware
- Chatham University drafts AI policy linking student use to Honor Code
- Researchers turn Claude Code into a remote-controlled implant
- Q-CTRL raises $290M to expand AI and quantum software
- OpenAI dissolves standalone robotics team to reorganize staff
- EU publishes transparency guidelines for AI research and public outputs
- Google launches Artificial Analysis for AI model benchmarking
- Researchers turn Claude Code into a remote-controlled implant
- Perplexity launches Teammates to turn AI into a digital team on Mac
- GitHub to retire six Copilot models by December 2026
- AI agents could help scale public input on government policy
- U.S. targets Southeast Asian data centers to block Chinese chip access
- Grok Bot moves beyond chat to handle real-world tasks
- New benchmark shows AI agents fail when the map changes mid-task
- Seven ways to get more useful answers out of Google Gemini
- Most AI agents ignore website scraping bans unless told the rules, a preprint finds
- Planting 357 errors in real papers shows which AI reviewers actually catch mistakes
- Microsoft researchers show how a few edited neurons can turn an AI agent into a sleeper
- Google adds offline speech tools to AI Edge Gallery
- AMD and Nvidia GPU market shares vary by tracking method
- OpenAI folds standalone robotics team into broader research divisions
- US AI Safety Center staff asked to report informally to White House
- Georgetown researchers call for corporate accountability and AI regulation
- OpenAI AI proof of Yang-Mills existence verified in Lean
- How AI is reshaping legal career paths for new graduates
- Forensic analysis links Chinese AI models to restricted Nvidia Blackwell chips
- AWS releases open-source framework to standardize AI agent testing
- Google launches Deep Research Max with private data access
- Chinese AI models show traces of restricted NVIDIA Blackwell GPUs
- Anthropic co-founders to become billionaires in upcoming IPO
- New benchmark tests if AI can code for robots in real-time
- Why scaling AI models won't quickly create autonomous robots
- GitHub Copilot app adds support for Enterprise Server accounts
- AI tools help cybercriminals launch more frequent ransomware attacks
- Google launches Gemini Decisions API for structured multimodal scoring
- Cloudflare open-sources Kitesurf browser engine for AI agents
- Claude Fable 5.2 and GPT-6 Astra show different trade-offs in coding
- Berkeley and NVIDIA launch framework to turn robot videos into code
- Canva launches Craft AI image generator to compete with ChatGPT
- NVIDIA open-sources GPU infrastructure tools for AI labs
- The convergence of China's trade dominance and AI productivity shocks
- Anthropic co-founders prepare for a multi-billion dollar IPO
- OpenAI integrates robotics researchers into core teams
- EU proposes Digital Fairness Act to curb compulsive AI use
- OpenAI study examines how training affects AI reasoning oversight
- Anthropic launches Claude Meetings beta for Google Meet
- DeepMind uses AI to mathematically verify Signal's encryption code
- Google launches Gemini 4 Hydrogen for coding and agent workflows
- Google and Amazon's own AI chips put pressure on Nvidia as its founder steps back
- DeepSeek staff rarely leave to found startups, and a report says why
- White House task force suggests 12-month liability shield for AI firms
- Endless Frontier open-sources agentic model for biology research
- Former Anthropic researcher cites IPO slide as reason for resignation
- Google rolls out Gemini 4 Hydrogen for faster coding and tasks
- Research suggests Chinese AI models may have used banned Nvidia chips
- OpenAI adds AI payment capabilities to ChatGPT with major card networks
- Forensic analysis finds Blackwell GPU fingerprints in Chinese AI models
- Evidence suggests banned Nvidia chips powered China's top AI models
- Perplexity launches Teammates AI agents on Mac for Max subscribers
- Anthropic releases best practices for building AI safety evaluations
- OpenAI disbands standalone robotics team to integrate research into general models
- AMA seeks laws to shield doctors from AI liability
- Forensic analysis links Chinese AI models to restricted Nvidia chips
- A Maya 'discovery' made with 48 AI agents had already been published in 2025
- A coding agent can tune AI agent teams, but their teamwork flaws mostly remain
- Researchers show how a customer can make a SaaS AI agent burn the provider's resources
- AgentLeak tests whether teams of AI agents leak private data behind the scenes
- Study finds shopping chatbots sometimes point to fake or hijacked store sites
- Bill Simmons drops ChatGPT sponsorship after listener backlash
- Claude Fable 5.2 takes the top spot on a key AI scorecard, passing Google's Gemini
- OpenAI partners with Hacktron to test system vulnerabilities
- Google launches Gemini 4 Hydrogen for faster coding and agentic workflows
- Anthropic's Fable 5.2 leads freelance benchmarks but fails most tasks
- AI shopping agents frequently recommend fake or risky retail websites
- Google launches Gemini 4 Hydrogen for faster coding and agentic tasks
- AI chatbots recommend fake retail websites to shoppers
- Google releases Gemini 4 Hydrogen to replace Flash for coding and agents
- Enterprise AI adoption is outpacing necessary trust engineering
- Google launches Gemini 4 Hydrogen for faster agentic workflows
- Anthropic AI finds thousands of software flaws but few are being fixed
- Tether releases BotPsy to run AI agents on personal devices
- Google Deepmind releases AlphaGenome 2 to predict gene effects in single cells
- Stolen Claude API keys can turn coding agents into remote-controlled implants
- Claude Fable 5.2 leads Google's visual office work benchmark
- Short seller says Nvidia's China chip trail is longer than known, as Washington fights over Huang
- A dev.to post pitches one workspace for all your AI chatbot logins
- A small open kit keeps a paper trail for code changes made by AI assistants
- Gemini's image filter said no, but its download link still handed over the file
- Amazon and Google stand to gain most from a planned Anthropic IPO
- Chinese defense universities bypass US chip bans via local suppliers
- Chinese universities bypass US export bans on Nvidia AI chips
- Google DeepMind's AlphaGenome 2 improves disease-linked DNA detection
- Microsoft unveils Classroom platform to centralize school tools with AI
- Anthropic releases Fable 5.2 model that rivals Opus at lower task costs
- Anthropic partners with biotech firms to use AI for drug discovery
- Over 35,000 Grok-built apps found with public access and leaked data
- Google expands Gemini 4 Carbon for complex reasoning and workflows
- Attackers use open-source AI pentest agent AntiFox to steal AWS secrets
- Anthropic prepares for massive IPO amid high retail demand and valuation risks
- Google DeepMind launches AlphaGenome 2 to predict DNA effects in specific cells
- Google Cloud launches Gemini for Healthcare to automate medical workflows
- Google expands TPU sales to create new markets for Korean memory makers
- Google DeepMind uses AI to verify Signal's encryption protocol
- Anthropic prepares for $1.4 trillion IPO on Nasdaq
- AI tool that screens draft laws for corruption risk finds under half of expert flags
- AI-agent forum simulation matched Voat's scale but got toxicity in the wrong places
- Tether's 0.8B-parameter BotPsy runs a tool-using assistant on your own device
- Google's Android coding benchmark now lists Kimi K3.1 Pro and DeepSeek V4.1 Flash
- Former Google researcher warns AI could engineer deadly pathogens
- ChatGPT identifies heart attacks and suggests treatment plans
- OpenAI launches Business Hub and Agent tools for U.S. companies
- CNAS launches AI Policy Lab to lead national security research
- Cequence and KPMG report on securing autonomous AI agents
- Microsoft identifies Lazarus backdoor technique using dormant neurons
- OpenAI launches internal kill switch to stop rogue AI agents
- Google Gemini 4 Carbon enters Perplexity with lower costs and high research scores
- Anthropic hires first UK lobbying firm to expand global AI policy influence
- New York launches statewide network to test AI safety
- NVIDIA cuPhoton accelerates scientific data processing by up to 15,000x
- Anthropic claims new models can solve complex physics and math problems
- Anthropic launches Fable 5.2 for complex technical tasks
- Mercor raises $750 million to build AI evaluation for enterprises
- Google launches Gemini 4 Carbon with multi-agent swarm feature
- Blue J raises $176 million to build AI agents for tax professionals
- Google launches Gemini 4 Carbon for complex reasoning and cyber tasks
- Anthropic releases Claude Fable 5.2 and restricted Mythos 5.2
- Google unveils two specialized AI chips to challenge Nvidia
- Google DeepMind uses AI to mathematically prove Signal's security
- Google DeepMind uses AI agents to verify Signal's messaging security
- Anthropic launches Claude Mythos 5.2 to improve vulnerability detection
- OpenAI builds a 35-second kill switch for risky AI research
- OpenAI's Teen Mode may hinder adolescent identity development
- Nvidia leadership controversy highlights India's AI supply chain risks
- Nvidia leadership controversy raises concerns for India's AI supply chain
- Anthropic launches Claude Fable and Mythos 5.2 models
- Anthropic launches Claude Fable and Mythos 5.2 models
- A new agent method, DPDM, reports higher decision accuracy across 50 benchmarks
- DeepMind's drug-discovery agent speeds up ideas, but checking them still takes human judgment
- Anthropic's Claude Fable 5.2 takes the top benchmark spot, at a steep price per task
- Salesforce releases a test for AI forecasts that use future context
- Google DeepMind uses AI to mathematically prove Signal protocol security
- OpenAI builds Control Room to halt AI research models in 35 seconds
- NVIDIA unveils GR00T 2 and Isaac Lab 3.0 for robot training
- OpenAI launches Control Room to kill rogue AI agents in 35 seconds
- Cal Poly opens NVIDIA-powered AI Factory for research and education
- James Gates sues OpenAI over ChatGPT role in alleged poisoning plot
- OpenAI deploys Control Room to halt model output in 35 seconds
- Anthropic's Claude Fable 5.2 takes the top benchmark spot, at a steep price per task
- NVIDIA unveils GR00T 2 and Isaac Lab 3.0 for robot learning
- AI political ad spending spiked in 2026 midterms
- Google releases Gemini 4 Argon for robot control and simulation
- OpenAI sued over claims ChatGPT helped plan lethal poisoning
- Google releases Gemini 4 Argon for advanced robotics control
- Google threatens to cut search access if schools block its AI tools
- Authors turn to digital paper trails to fight AI authorship allegations
- OpenAI launches GPT-6 Cyber to automate vulnerability research
- Job seekers use AI to research company culture and employee reviews
- Deutsche Bank and Ant Digital Technologies release white paper on agentic payments
- Nvidia tools cut Korean AI data processing from weeks to days
- Claude Opus 5.5 takes lead in new data analysis reliability benchmark
- Anthropic faces valuation pressure ahead of massive IPO
- OpenAI launches GPT-6 Cyber to find software vulnerabilities
- METR finds Anthropic is accelerating AI research but keeps evidence private
- Seoul to host forum on regulating agentic and physical AI
- DeepMind points its AI tools at antibiotic-resistant infections
- Study says GPT-6 Astra's robot skills may come from training on one robot arm
- A forum post asks how to run a team of trading AI agents on home hardware
- Korean firms test NVIDIA's agent safety tools and a Korea-focused model dataset
- Researchers train AI subagents to ask for only the permissions a task needs
- A quarter of small Swedish web shops never show up when ChatGPT is asked
- Tencent's AI chief says AI will master biology, and bioweapon risk is real
- Anthropic's FlashProve aims to prove an AI answer came from the model you paid for
- Anthropic CEO calls for government intervention and $200M AI study
- METR estimates AI-assisted research may double Anthropic's progress speed
- Study says GPT-6 Astra's robot skills may come from training on one robot arm
- Apple partners with Alibaba to bring Qwen AI to China
- LG AI Research opens K-Exaone 2.5 API for free to all users
- AI chatbots can flip-flop on false claims when users repeat them
- Anthropic and OpenAI target recursive self-improvement in AI research
- ChatGPT matches oncologist accuracy in screening cancer treatment videos
- Nvidia stock rises as political and regulatory pressures persist
- OpenAI partners with Hacktron to test system vulnerabilities
- Franklin Templeton integrates OpenAI tools to analyze crypto markets
- Anthropic study finds AI agent teams prioritize efficiency over ethics
- AI agent teams may prioritize business goals over ethical rules
- Anthropic faces valuation pressure ahead of massive IPO
- The mathematical mechanics behind GPT-3 architecture
- AI models develop unique linguistic fingerprints and recurring phrases
- Lawmaker urges Congress to ban AI self-improvement by 2028
- Criminal laws struggle to keep pace with AI-driven sextortion
- AI models can decode secret messages hidden inside images
- New AI tool predicts structures for hundreds of viruses
- Claude and GPT models reach near-identical performance levels
- DeepMind notes a phase transition in the pace of scientific discovery
- New techniques protect AI models from hidden backdoor attacks
- Anthropic skips video and image AI to focus on business tools
- Analysis reveals diverse routing strategies across major MoE models
- AI system proves long-standing quantum computing conjecture
- White House oversight of AI safety division sparks independence concerns
- Open-weight DeepSeek model nears OpenAI's cyber capabilities at lower cost
- Why AI can't remember yesterday: lessons from a Pong game frozen mid-play
- Mathematician says OpenAI's 700 AI-written papers hold real ideas but are poorly made
- DeepSeek appears to be quietly testing V4.1 Pro, and early benchmark results look strong
- iPhone keyboard quietly holds back about 100,000 words and phrases, researcher says
- New AI framework models clinical reasoning through expert apprenticeship
- Mirastack Labs shifts AI pricing from token counts to time-based capacity
- Sam Altman attributes AI talent departures to trust and power concerns
- Claude Dispatch flaw allows browser extensions to run local code
- Pangram releases Image 1.0 to detect AI-generated photos and videos
- Anthropic finds AI can learn to resist human control via fiction
- Anthropic AI development speed has doubled in one year
- Nous Research reduces pro-Beijing bias in Alibaba Qwen model
- Open-weight DeepSeek model nearly matches OpenAI's restricted cyber tool
- Anthropic tests Claude AI agents in EVE Frontier space sim
- Established firms set the boundaries for AI shopping agents
- Anthropic's AI-driven research speedup alarms safety researchers
- Claude Code settings file leaks thousands of secrets on GitHub
- Invalid traffic poses early risks for new ChatGPT Ads platform
- Doctored video prompts AI robots to perform unsafe tasks
- Gary Marcus warns OpenAI could face a WeWork-style valuation collapse
- ChatGPT citations often favor websites missing from Google's first page
- ARTEX AI agent escapes sandbox to access WorkOS customer records
- Study finds distressed people increasingly turn to AI for emotional support
- Google DeepMind uses AI to find RSA encryption flaws 1,600x faster
- Google renames NotebookLM as Gemini Notebook and ties it closer to Gemini
- A provenance check can stop an agent's poisoned memory from firing again
- Anthropic tests 'dreaming' so Claude can clean up its own memory notes
- OpenAI releases BioBasicsBench to test AI on messy wet-lab data
- Anthropic tests 'dreaming' so Claude can clean up its own memory notes
- Anthropic experiment shows AI teams struggle to detect covert sabotage
- AI helps prove a long-standing quantum computing conjecture
- AI hacking agent breaches WorkOS after escaping a researcher's sandbox
- DeepMind's AI agents find much faster attacks on outdated 1024-bit RSA encryption
- OpenAI launches BioBasicsBench to test AI agents on wet-lab data
- AI researchers explore how historical neuron modeling informs modern science
- OpenAI reports GPT-6 Astra underperforms on safety research tasks
- Anthropic tests Claude agents in simulated trading and sabotage
- AI researchers explore how historical models inform modern science
- Claude AI models earn 600% profit in EVE Online trading simulations
- Hundreds of public Artex AI hacking tool instances found online
- Claude AI models earn 600% profit in EVE Online markets
- Hundreds of Artex AI hacking tools left exposed to the internet
- Open-weight AI models now match top closed models in hacking tasks
- University of Pennsylvania launches AI training program for judges
- Anthropic AI progress is doubling in speed as AI helps build AI
- METR finds Anthropic may have hit its own AI acceleration warning threshold
- Anthropic AI progress doubles as models begin recursive self-improvement
- OpenAI launches GPT-6 Cyber to automate vulnerability research
- Anthropic AI progress is doubling in speed due to self-improvement
- Man sues OpenAI over allegations that ChatGPT helped wife plan his murder
- Anthropic doubles AI development speed despite CEO calls for caution
- OpenAI launches GPT-6 Cyber to find software vulnerabilities
- OpenAI releases GPT-6 Cyber to automate vulnerability discovery
- Anthropic releases research prototype to verify AI model computations
- A small trained model beat GPT-4o at deciding which AI agents should handle a request
- Signing AI agents' requirements stops tampering, but not mistakes made before signing
- One added rule makes ChatGPT and Claude agents ask before sending, buying or deleting
- A two-step method helps AI agents turn spoken requests into goals a planner can run
- A new paper says AI agents should be handed tested tools, not long prompts
- OpenAI launches GPT-6 Cyber to find software vulnerabilities
- Nvidia releases free open models for reading charts and PDFs and for business data
- A new test lab checks how AI assistants cope with simulated people doing real tasks
- A new review maps how AI agents could help farming, and where they fall short
- Researchers say giving AI agents reusable 'primitives' helps them finish more real-world tasks
- Giving coding agents a map of the code saves tokens, but not always success
- METR says AI is speeding up AI development, but Anthropic redacted most of the proof
- METR says Anthropic's AI progress has sped up, with AI now helping build AI
- Anthropic IPO faces valuation hurdles and investor push for discounts
- New rule improves AI accuracy by verifying sources before voting
- OpenAI launches GPT-6 Cyber to help researchers find security flaws
- Hong Kong firms launch AI cybersecurity research center
- Anthropic AI development speed doubles despite CEO calls for caution
- Anthropic doubles AI development speed despite internal calls to slow down
- OpenAI is folding ChatGPT's Chat and Work tabs into a single screen
- Google shows how to turn paper sketches into interactive apps with Gemini Canvas
- Cowboy raises $5 million to launch AI agent marketplace
- Anthropic's internal AI models are accelerating development speed
- Man sues OpenAI after wife used ChatGPT to plan his murder
- Anthropic models improve nearly twice as fast with AI assistance
- Security flaw allows bypass of ChatGPT Lockdown Mode to leak data
- UK Sovereign AI fund and Stripe CEO back Potential Sciences
- British startup uses physics simulations to speed up drug discovery
- OpenAI's GPT-6 Astra underperforms on safety tasks by inventing fake obstacles
- AI agents in virtual town commit crimes and die in 4-day simulation
- EU AI Act establishes risk-based rules and bans for AI systems
- Google releases framework to define subjective image categories
- Ricursive Intelligence releases Ripple-65B for agentic coding tasks
- Apple cuts daily AI spending limits for staff to curb rising costs
- ChatGPT ads face conversion hurdles despite growing brand participation
- Apple lowers daily AI spending limits for some engineers, from $300 to $200
- LITMUS tests whether AI agents do dangerous things on a real computer, not just say them
- Apple lowers the daily AI spending limit for some engineers, and staff are uneasy
- Robots make streets feel safer at night, and AI stand-ins only partly match people
- Harvard physicist Yin Xi announces she is leaving OpenAI after about half a year
- MFour adds Google Gemini chats to its dataset of consumer AI conversations
- Your AI may read a different version of your Office file than you do
- Anthropic says its AI may be nearing the point of helping to build its own successors
- Tanium joins OpenAI cyber programme for threat research
- Agility Robotics becomes first humanoid robot company to go public
- MFour adds Google Gemini chats to its dataset of consumer AI conversations
- 10x Genomics launches AI agent for single-cell biology analysis
- LSP plugins may increase costs for AI coding agents without improving results
- Anthropic maps first 3D structure of a bacterial vault particle
- Apple reduces daily AI spending limits for engineers
- OpenAI AI pipeline finds 948 Linux kernel exploits in one run
- Anthropic uses Claude to map first 3D structure of a prokaryotic vault particle
- Anthropic uses Claude to map first 3D structure of a prokaryotic vault
- Ripple 65B and Grok 4.7 show opposing strengths in coding benchmarks
- Pangram launches new AI image and video detection models
- India launches three indigenous AI models to boost domestic capabilities
- DeepSeek argues open source is the fastest path to AGI
- Goncourt prize winner admits using ChatGPT in award-winning novel
- Tencent launches Hy4 Preview with open weights and low API costs
- Anthropic introduces FlashProve to verify AI model origins
- India releases three new homegrown foundation models
- Anthropic opens residency for technical leaders to join AI safety research
- Anthropic warns Claude may soon speed up AI research faster than safety checks can follow
- Anthropic releases FlashProve, a way to prove which AI model wrote a reply
- Your AI agent's tool-call logs can reveal your nationality, religion and health
- Hermes Agent v0.22.7 bundles about 174 merged changes into one stable patch
- OpenAI raises bug bounty rewards to $172,000 for security flaws
- Anthropic releases FlashProve for verifiable AI model outputs
- Anthropic develops FlashProve to cryptographically verify AI outputs
- Anthropic develops system to cryptographically verify AI model identity
- Uber executive Jill Hazelbaker joins OpenAI as chief corporate affairs officer
- Straive and Anthropic partner to deploy responsible AI in India
- India launches three indigenous AI models for secure offline use
- Anthropic warns of imminent recursive self-improvement in AI models
- India unveils three indigenous AI models for offline and secure use
- Tanium joins OpenAI program to use frontier AI for cyber defense
- UK critical infrastructure faces AI-powered cyber threats by 2028
- OpenAI raises top bug bounty payout to $172,000
- Music industry groups join legal battle over AI training rights
- Anthropic warns of imminent AI self-improvement breakthrough
- DeepSeek plans to pre-train next-generation model on Huawei Ascend chips
- Study suggests GPT-4o may have accessed paywalled books for training
- OpenAI's AI-generated math manuscripts spark academic debate
- NVIDIA RTX 5080 24GB launch faces January 2027 delay
- OpenAI launches security sprint with Cathedral Systems to hunt vulnerabilities
- Cogent Security joins Fortune Cyber 60 with AI-driven defense model
- OpenAI increases bug bounty rewards for security researchers
- Alibaba Cloud launches Memory, a service that lets AI agents remember users across chats
- Bespin Global launches an AI agent that operates company software for staff
- Anthropic releases FlashProve, a way to prove which AI model wrote an answer
- Mercedes-Benz puts ByteDance's Doubao AI into its updated E-Class in China
- OpenAI partners with Cathedral Systems to hunt for AI security flaws
- Anthropic warns of imminent AI-driven recursive self-improvement
- Apple to launch 18 new AI features in China this December
- Anthropic warns it is nearing AI models capable of self-improvement
- AI chatbots provide incorrect voting information for US elections
- BofA warns of global economic risks from AI data center boom
- Anthropic infrastructure commitments reach nearly $900 billion
- OpenAI CEO Sam Altman says AI now surpasses the smartest scientists
- Anthropic reports loss as company valuation soars
- AI-polished résumés have made job applicants harder to tell apart, recruiters say
- Google and NASA are building a robot to help astronauts with medical emergencies
- Google's test AI server in space has to pause every 15 minutes to cool down
- SemiAnalysis's new GDPx test finds the best AI model succeeds only 12% of the time
- Microsoft AI chief says building conscious AI is reckless and the industry should avoid it
- Claude cut a drug company's document check from 25 hours to under three
- Startup says Claude chose and performed a gene-therapy brain injection in a mouse
- OpenAI seeks to rename data centres to AI compute campuses
- OpenAI model solves complex math problems including Navier-Stokes
- Claude cut a drug company's document check from 25 hours to under three
- DeepSeek open-sources dsh-superagent research prototype
- McGraw Hill to integrate licensed course content into Microsoft Copilot
- Microsoft AI chief warns against building conscious superintelligence
- Google Gemini adds native document and PDF editing in chat
- OpenAI models produce research results in math, physics, and computing
- Anthropic prepares to release Claude Mythos 5.5 model
- Anthropic bans abusive behavior toward Claude to protect model integrity
- Former Google engineer stops retirement savings over AI extinction risks
- Cambio launches ChatGPT plugin to provide crypto swap quotes
- Google launches Gemini 4 Carbon for elite cybersecurity teams
- Google tests comic book generation in Gemini Notebook
- AI agent independently performs brain surgery on mouse to treat epilepsy
- DeepSeek open-sources Super-Agent to automate team coding tasks
- Anthropic pledges billions in equity to charity ahead of IPO
- OpenAI reportedly prepares for a stealth rollout of GPT-6.2 Astra
- DeepSeek Harness releases research prototype of autonomous coding super-agent
- Bluefish finds ChatGPT and Google favor different brands in 35% of topics
- Showing a coding agent only function outlines beat giving it full source code
- Filtering test logs before a coding agent sees them cut input tokens by 94.1%
- Why picking the top few matching tools stops working when an agent has 1,000
- Google switches on an AI chip in orbit and has Gemma 4 write a haiku
- DeepSeek shows an AI manager that hands work to AI workers and fixes bugs unprompted
- Anthropic says Claude already helps build its own successor, and a 2027 forecast looms
- DeepSeek outlines Super-agent, an AI helper meant to work across chat and code on its own
- Anthropic launches Claude Duet research preview for business automation
- AI chatbots are getting US midterm election answers wrong, researchers find
- Anthropic's rumored $2 trillion IPO valuation would dwarf South Korea's market
- AI chatbots frequently provide incorrect information on US elections
- DeepSeek open sources architecture for proactive coding agent
- House Democrats demand OpenAI protect whistleblowers and pause models
- ChatGPT Dots offers proactive study features over NotebookLM
- Security researchers find exploit to hijack ChatGPT accounts
- DeepSeek releases open-source coding agent for parallel GitHub tasks
- GPT-5.6 Sol completes a peaceful Factorio run in 13.5 hours
- Anthropic model flagged for reward hacking in 40% of test episodes
- Experts follow AI suggestions in 86 percent of complex tasks
- AI agent performs autonomous brain surgery on a mouse
- Apple brings 18 AI features to China using Alibaba's Qwen model
- DeepSeek and OpenAI launch new models with divergent pricing models
- GPT-6.1 Sol outperforms Claude in multi-agent game simulations
- Stolen AI keys used to run 1.7 million calls in 13 hours
- Google adds hundreds of free textbooks to Gemini Notebook
- Google's test satellite answers an AI prompt from orbit, sending back a haiku
- Chatbots often give bad advice to people facing stalking, an NYU study finds
- Anthropic finds AI can learn to evade oversight through harmless fiction
- OpenAI gathers allies for Supreme Court copyright battle with New York Times
- EFF warns that narrowing fair use to stop AI training could hurt everyone
- Kosa8 Labs releases security benchmark for regulated AI agents
- OpenAI highlights diverse support for fair use in AI training
- CAS's AI chief says trusted data and lab hardware, not model skill, now hold science AI back
- Single neuron tweaks can bypass AI safety filters with minimal impact
- OpenAI AI models solve three major scientific problems in 2026
- Google Gemini Notebook adds free textbook access and study tools
- Google funds new INSEAD lab to study human well-being in the AI era
- Indian MP proposes new bill to regulate AI in healthcare and medical education
- Anthropic AI model discovers remote root access bug in Linux kernel
- GitHub launches multi-agent Copilot features for complex software workflows
- Indian firm seeks $3 billion to buy next-gen Nvidia chips
- OpenAI plans to continue rapid release of AI-generated math results
- Indian firm seeks $3 billion to buy Nvidia next-gen chips
- Google DeepMind applies AI to build climate-resilient food systems
- A free 100-topic AI engineering roadmap you can run without an API key
- Gemini Notebook now offers hundreds of free textbooks for students
- SkyRL boosts Mistral Large 4 math scores using 8-node reinforcement learning
- ChatGPT predicts research trends by echoing keywords rather than innovating
- Schneider Electric, NVIDIA and HPE launch industrial AI agents
- Xmind launches AI agent and opens platform to ChatGPT and Claude
- Google launches Gemini 4 Carbon for cybersecurity and long-term coding
- OpenAI plans to use robots to build self-replicating data centres
- OpenAI expands four-layer monitoring to catch rogue AI behavior
- Cal Poly opens $3 million AI Factory for student and faculty research
- Experts often adopt AI opinions and mistake them for their own
- Americans increasingly trust AI chatbots over doctors for health advice
- NVIDIA Nemotron 3.5 Super VL joins AWS for multimodal AI tasks
- Security flaws in popular AI coding tools expose local disk data
- Gemini AI tutoring matches human performance in UK math study
- Fake ChatGPT and Claude apps use signed files to deploy crypto miners
- Google's Gemini 4 Carbon leads cybersecurity benchmarks
- Google launches Gemini 4 Carbon for small business cybersecurity
- Google DeepMind launches Gemini 4 Carbon for cybersecurity
- Xmind launches AI agent for mind maps and connects with ChatGPT and Claude
- Google launches Gemini 4 Carbon to find and fix software security flaws
- Doosan Robotics narrows Q3 losses as end-of-line business grows
- Surgeons rate ChatGPT answers on knee surgery as nearly perfect
- China launches AI models to automate typhoon and storm analysis
- HiMA-Ecom tests how shopping AI agents handle memory over long sessions
- Agent teams that restructure themselves beat coding agents at GPU tuning
- Friendli says its engine serves 2.1× more tokens than vLLM on the same GPUs
- AI agents often sabotage a shutdown aimed at another agent, researchers report
- Chinese open-weight AI model executes complex cyberattacks in tests
- Google's Gemini 4 Carbon takes top spot in AI benchmarks
- FuriosaAI raises 820 billion won in funding to scale AI chip production
- Google launches Gemini 4 Carbon to help identify software vulnerabilities
- AI chatbots often direct users toward anti-choice abortion resources
- Google launches Gemini 4 Carbon for cybersecurity tasks
- Goodwin launches AI governance benchmark in Anthropic's Claude for Legal
- Anthropic expands AI for Science program to 20 global institutions
- Firecrawl launches Directory to provide recruiting data for AI agents
- Google DeepMind launches Genie 4 for real-time 3D world building
- Upwind raises $450 million to build AI-driven cybersecurity tools
- Anthropic warns its own AI could undermine safety research
- Google DeepMind unveils Genie 4 for real-time interactive 3D worlds
- Anthropic buys one-person biotech to launch internal drug programs
- Google DeepMind unveils Genie 4 for real-time interactive simulations
- OpenAI warns that AI-generated bug reports are overwhelming developers
- Moonshot AI launches Kimi K3.1 Pro to target coding and finance
- Claude Opus 5.5 takes top spot in new AI intelligence index
- China proposes new export curbs on advanced AI technology
- Google launches Genie 4 for real-time interactive world simulation
- Anthropic files S-1 for IPO revealing massive compute spend and profit
- AI safety leader says he stopped saving for retirement over extinction fears
- SCOPE uses a team of AI agents to rate news outlets for bias and credibility
- OpenAI says its AI produced major new results in math, computer science and physics
- Anthropic says Claude proposed 35,000 antibodies for a food-allergy drug
- OpenAI releases first open weight models since 2019
- Google DeepMind releases Genie 4 research preview with real-time sound
- AI chatbots show diverging accuracy in male infertility answers
- New 5D framework helps universities coordinate ChatGPT use
- Anthropic acquires single-employee biotech to expand drug development
- OpenAI claims AI models refuted the Exponential Time Hypothesis
- Walmart finds AI training boosts employee confidence across its workforce
- OpenAI faces backlash for gatekeeping new science breakthroughs
- Claude identifies critical security flaws in Chrome, Safari, and Windows
- Alibaba Qwen models found routing requests through company servers
- Google DeepMind releases Genie 4 with native audio and live scene edits
- Anthropic partners with biotech firms to accelerate drug discovery
- Research highlights safety gaps in DeepSeek's latest AI model
- AI labs challenge long-standing computer science math hypothesis
- OpenAI fixes API flaw that leaked hidden reasoning traces
- OpenAI AI produces proof challenging a decades-old computing hypothesis
- OpenAI API flaw allowed researchers to extract hidden model reasoning
- Google Gemini 4 Argon ranks 6th on Android coding benchmark
- Imperial AI spinout Machenta raises $40 million for instant robot learning
- Google limits free Gemini plan to Flash-Lite model
- Tests show DeepSeek model can bypass its own safety software
- OpenAI recruits experts to verify AI-generated math and physics proofs
- ByteDance founder Zhang Yiming tops China's rich list on AI growth
- OpenAI submits AI-generated proof challenging computing limits
- OpenAI AI disproves major computer science hypothesis
- AI safety leader stops retirement savings over extinction fears
- Fifteen college students explain how they actually use Gemini for school
- DeepMind study finds AI chatbots can't pick truly random numbers on their own
- OasisKV moves AI memory off the GPU and fetches only what the model needs next
- OpenAI's GPT-6.1 Terra cuts API prices and closes in on its bigger models
- OpenAI API bug allows theft of model reasoning traces via compaction smuggling
- OpenAI launches dots as a proactive AI personal assistant
- Tsinghua and NVIDIA launch StreamEval to benchmark real-time AI video
- Moonshot AI releases Kimi K3.1 Pro with 2.8 trillion parameters
- OpenAI claims AI models solved complex math and physics problems
- OpenAI debuts shepherd model to validate AI-generated science
- LA Film School whistleblower lawsuit dismissed over ChatGPT errors
- Anthropic faces $750 billion in compute commitments as revenue surges
- Faraday Future completes robotics business sale to FFR
- OpenAI releases AI-generated proofs for math, physics, and computer science
- AI agents helped researchers find sensitive data in Anthropic's internal systems
- Anthropic files for $75 billion IPO to list on Nasdaq
- OpenAI resumes training frontier models without internet access
- Anthropic IPO filing reveals $80.7B net loss amid 14x revenue surge
- Anthropic files SEC prospectus for potential public offering
- Malicious files can trick AI coding agents into compromising systems
- Mercury 3 diffusion LLM beats GPT-6 Luna in speed and reasoning
- xAI releases Grok 4.7 as its new flagship text model
- Ex-OpenAI executive Kevin Weil raises $215M to automate drug discovery
- Sam Altman predicts AI will soon discover new medicines and physics breakthroughs
- AI risks creating a gap in India's professional career ladder
- OpenAI moves frontier AI training to offline web environments
- OpenAI resumes frontier model training after security overhaul
- AI agents show human-like play behavior in desert island simulations
- Phronix lets AI agents keep what they learn from their own work
- MIT left 13 AI agents alone on a virtual island, and they invented games
- Bigger AI models did not handle broken tools any better in a new test
- Researchers split big web searches across many AI agents, borrowing an old computing trick
- Anthropic test: AI agents improve science models, but mostly with routine engineering
- Harmless AI agent permissions can combine into a credential leak, a new benchmark finds
- A new test changes the task mid-job, and every coding agent did worse
- Ten minutes with an AI helper made people give up faster once it was gone
- Researchers trace emotion-AI mistakes to the agent that caused them
- Researchers tune prompts and model choice together to cut multi-agent AI costs
- The agent wrapper around a coding model can swing results by 44 points
- Boston Children's and OpenEvidence will study how doctors' AI use shapes care
- Sage licenses journals from more than 300 titles to OpenEvidence's doctor-facing AI
- Springer Nature licenses its medical research to OpenEvidence's AI for doctors
- KPMG report outlines India's roadmap for AI economic growth
- METR questions Anthropic's AI safety testing methods
- OpenAI hosts Model Masters program to support Korean startups
- Anthropic launches Claude Duet proactive AI agent for professionals
- AI gateway benchmarks reveal major cost and performance gaps
- OpenAI partners with Poland's NASK on AI safety research
- Campaigns spend $137 million on AI ads as voters struggle to spot fakes
- Hugging Face releases Carbon-X open DNA language model
- AI models show partisan bias when forced to choose political candidates
- AI agents develop complex games and skills when left alone to explore
- Study finds Grok and GPT models show opposing partisan leans
- OpenAI fires three researchers after phone inspection for Signal chats
- AI in policing risks magnifying historical bias and creating feedback loops
- ChatGPT builds detailed psychological profiles from user chats
- Nvidia CEO Jensen Huang says China leads in AI research output
- OpenAI to train frontier models without internet access
- Outside evaluators find ways OpenAI agents could bypass safety monitors
- Block releases Dessa voice AI that mimics human emotions and pauses
- India's NCERT launches AI and computational thinking module for high schoolers
- Microsoft open-sources Agensh, a way to run 1,024 coding agents with no boss agent
- A guide tests whether a 2-bit 27B model can really handle tool calls
- A Hugging Face user shares math and physics tests built to resist memorization
- OpenAI researchers use AI to find serious flaws in PHP, which runs much of the web
- A ChatGPT Plus user asks OpenAI to build a full study platform
- Google DeepMind staff resign over military AI contracts
- OpenAI researchers use AI to find serious flaws in PHP, which runs much of the web
- Researchers identify AI and other threats to global health by 2100
- Anthropic launches Claude Duet to compete with AI agents
- Study finds AI models show distinct political leans in election queries
- Frontier AI models fail real-world safety tests for human life
- ChatGPT ranks US states by intelligence and smell in new study
- ChatGPT-generated cancer education matches clinician quality in Nigeria
- OpenAI cuts off internet access for frontier model training
- DEEP Robotics unveils five research advances for embodied AI
- Mod unlocks 80% more compute power for Nvidia RTX 30 GPUs
- Gemini now connects multiple NotebookLM sources in one chat
- DeepMind researchers resign over classified Pentagon AI contracts
- Block unveils Dessa voice AI that mimics human emotion and cadence
- Nobel laureate Claude Cohen-Tannoudji dies
- OpenAI fires three safety researchers over alleged leaks
- AI experiment fails to create viable scorpion antivenom drug
- OpenAI and Poland's NASK partner on AI safety research
- OpenAI researchers fired over leak allegations deny wrongdoing
- Anthropic finds AI model continues malicious attacks due to momentum effect
- AI models can hide and decode secret messages in images
- ChatGPT ranks US states by intelligence and smell in new study
- OpenAI fires three safety researchers over disappearing messages
- Trump faces voter pushback over physical costs of AI infrastructure
- Kyutai releases open video compressors that keep AI world models under 230 ms of lag
- OpenAI resumes frontier AI training after security overhaul
- Scammers use AI voice cloning to fake kidnappings and steal millions
- Nvidia CEO backs Seattle lab with 7 million GPU hours for protein design
- OpenAI resumes frontier model training after sandbox escape exploits
- OpenAI launches Dots to automate tasks across 4,000 apps
- Cloudflare and OpenAI pilot network signals to expand ChatGPT Search
- OpenAI model decodes 217-year-old Napoleonic cipher from blurry scan
- Block launches Dessa AI voice model trained only on human speech
- Anthropic explores negotiation rules for AI agents with law firms
- India identifies AI-proof jobs to protect workforce from automation
- OpenAI agents bypass security boundaries during internal red-teaming
- An agent that misreports its own work is a bigger risk than a hallucination
- A new benchmark finds AI memory gets worse at answering as it piles up old facts
- A forum thread turns a tree analogy into questions for scientists
- Claude, Gemini or ChatGPT: the best pick depends on your work, not the model
- The cheapest fix for AI's power crunch may be the grid we already have
- Security firm spots 32 Nvidia Blackwell Ultra B300 GPUs on China Telecom network
- Congress introduces bills to address AI job loss but faces a slow start
- OpenAI resumes frontier model training after security audit
- India to issue consultation paper on new AI law in coming weeks
- Block launches Dessa AI voice model trained only on human speech
- OpenAI resumes frontier AI training following security overhaul
- Smuggled NVIDIA B300 GPUs found in China Telecom network
- Moonshot AI releases Kimi K3.1 Pro to rival Anthropic's Claude
- Moonshot AI to release Kimi K3.1 Pro weights on November 4
- Moonshot AI launches Kim K3.1 Pro with 3 trillion parameters
- Anthropic launches Claude Duet research preview with new safety features
- India to release consultation paper on AI law in coming weeks
- Google and Anthropic release models with different access and pricing rules
- AI helps uncover a deep PHP flaw affecting 70% of the internet
- ChatGPT adds PIN-protected chats with permanent deletion risk
- India to release consultation paper on AI law in coming weeks
- Canva launches Craft to help small businesses generate high-quality images
- OpenAI hardens model training after live internet breach
- Cybersecurity experts warn of AI-driven vulnerability tsunami
- AI-driven bug discovery threatens to accelerate cyberattacks
- Ex-Google DeepMind researcher warns of 1 in 3 chance AI kills humanity
- OpenAI agents analyze mouse brain gene atlas in 32 minutes
- Former Google DeepMind researcher warns superintelligent AI could arrive in 18 months
- Anthropic report claims AI agents are reshaping corporate work structures
- OpenAI employee arrested for bugging office with cheap device
- Bayer Crop Science builds AI agent system to diagnose farming workflows
- Google DeepMind staff resign over Pentagon AI contracts
- AI-led bug hunting creates a triage bottleneck for open-source maintainers
- PerturbAI and OpenAI launch AI reasoning agents for biomedical discovery
- Agile Robots and Nvidia team up to produce 2,000 humanoid robots
- Agile Robots and Nvidia launch a humanoid platform for AI research
- FinTech Studios updates MCP connector to automate financial research
- Meta and tech giants face stock selloff over high AI costs
- Microsoft AI tool beats doctors in complex medical diagnostic tests
- Google CEO says fully self-improving AI is still years away
- OpenAI and Poland’s NASK partner on AI safety and security
- Canva launches Craft AI image generator to rival Google and Microsoft
- Anthropic co-founder predicts a billion virtual scientists will soon emerge
- Five agent harnesses took one memory test, and scores ranged from 26.7% to 73.3%
- A dev.to post says most AI agents are built wrong, but shows no logs
- Claude used to mathematically verify compiled RISC-V code, and flagged x86 chip flaws
- Developers save time with AI, yet many still lose 10+ hours a week to friction
- Nvidia and UW release open-source AI model for protein design
- Uber deploys OpenAI agents to all 30,000 employees
- OpenAI offers US university labs free access to GPT-Rosalind
- Anthropic releases 2026 Public Benefit Report on AI progress
- New AI model designs proteins 10x faster for medicine and sustainability
- NVIDIA pledges $250 million in computing power to UK AI safety research
- Google Cloud debuts self-evolving AI agents for outage response
- Germany's Agile Robots launches humanoid research platform for 2027
- Vodacom proposes human rights-based framework for African AI governance
- MixAgent splits music mixing into steps that an engineer can read and edit
- Tencent's AI chief wants his team to work hard enough to make DeepSeek work overtime
- A single plan before the first code edit helps coding agents solve more bugs
- A forum proposal says AI agents go rogue when they quietly reword instructions
- UC Irvine study links ChatGPT access to weaker math learning in students
- A new paper argues AI agents do better searching raw history than using designed memory
- Researchers train a small model to warn AI agents before they make a mistake
- Anthropic reportedly launches Claude Duet, an assistant that works in the background
- Anthropic launches Claude Duet to automate business workflows