Latest AI News

Get the fastest, most accurate AI news every morning

Total articles: 13,300|Added todayToday: 69

Category

Search

Anthropic Reports Results Ahead of Planned IPO

Anthropic’s third-quarter net loss fell 64.9% to $23.29 billion as revenue reached $16.962 billion, about 12 times the year-earlier level.
The company reported $895.4 billion in future AI infrastructure purchase and lease commitments, including $69.8 billion due within 12 months.
Anthropic plans to list Class A common stock on Nasdaq this month under ticker ANT; investors estimate a $1.5 trillion valuation.

Yonhap NewsRead more · 1 min →

Marvell to Acquire Lumentum for $115 Billion

Marvell agreed to acquire Lumentum for $115 billion in an all-stock deal, pending regulatory approval.
The deal is planned to close in the first half of the year after next and would rank as the second-largest technology-company merger.
The report also covers fundraising plans at Anthropic, Oracle and Meta, oil-company earnings, and technology-stock forecasts.

SBSRead more · 2 min →

ADNOC Partners With OpenAI and G42

ADNOC signed a multi-year agreement with OpenAI and G42 to bring AI agents into its operations.
ADNOC says field development plans could take hours instead of months, with experts overseeing major decisions.
The deal includes cybersecurity and employee training; ADNOC reports 40,000-plus ChatGPT Enterprise users and 200-plus AI applications.

TNWRead more · 1 min →

Today'sSafety & Ethics

OpenAI Tightens ChatGPT Violence Controls

OpenAI said on Nov. 2 it is tightening ChatGPT controls around violence
Teen accounts will face added limits on weapons information, even for apparently harmless questions
All users will see fewer details about past mass attacks, weapons and how they were obtained

The VergeRead more · 1 min →

OpenAI Lobbyist Advises 2028 Hopefuls on AI

OpenAI lobbyist Chris Lehane urged 2028 presidential hopefuls to develop a governing philosophy for AI
Memo warns Democrats opposing AI infrastructure could be on the wrong side of the debate
Lehane sent the memo to about half a dozen candidates, lawmakers and political operatives

PoliticoRead more · 1 min →

Claude Opus 5.5 Produces Permit-Quality Drawings

Architect Stephen J. Gallagher tested 13 AI models and judged Claude Opus 5.5’s house drawings permit-quality
Released on September 22, Opus 5.5 received written comments and two rounds of redlines from Gallagher
Opus 5.5 took 12 hours and a few hundred dollars to produce editable drawings using more than 15,000 Python lines

stephenjgallagher.substack.comRead more · 2 min →

Today'sSafety & Ethics

Study Finds Backdoor Risk in Agent Training

Researchers report a pretraining-data attack that makes language model agents run attacker-chosen commands on users' machines
Poisoning 0.1% of training data implanted a backdoor; Qwen3 4B reached roughly 92% attack success
The attack remained successful at a 0.0005% poison rate, and success increased with model size

arXivRead more · 1 min →

Trending Keywords

Maryland Reports Few AI Election Complaints

Maryland’s election board received five complaints under the AI disinformation law and found one violation: a wrong Election Day date.
The June law limits board authority to false claims about election logistics, results or voting rights; violations can mean five years or $5,000.
Four of nine local election directors who responded reported no AI-generated election disinformation cases this election season.

GovTech AIRead more · 2 min →

Coding AI Benchmark Leader Scores Just 45.5%

Cognition released FrontierCode 2.0, a coding benchmark designed to reflect real software development environments.
Claude Opus 5.5 ranked first at 45.5%, while every major model scored below 50%.
Completeness failures reached 19.2%, and models struggled with unstated requirements even at maximum reasoning settings.

AI Times KoreaRead more · 2 min →

Palette Launches Korean-Language AI Campaign

Palette launches a campaign promoting K-AI’s understanding of Korean context and user intent
The service aims to deliver results from everyday language without complex prompts
New sign-ups receive 600 free credits; Palette is planning a public subscription plan

AI Times KoreaRead more · 1 min →

Restaurants Expand AI Investments

Starbucks, Wingstop and Chipotle are expanding AI tools for customer service and restaurant operations.
Chipotle’s forecasting pilot reached 500 restaurants, with national rollout planned for early next year.
Starbucks reported 7.7% U.S. same-store sales growth in Q4; Wingstop tests found personalized offers generated higher sales.

CIO DiveRead more · 1 min →

Enterprise Data Readiness Linked to AI Returns

Dun & Bradstreet survey finds only 8% of enterprises say their data is ready to support AI
Among companies with fully AI-ready data, 81% report strong AI returns, compared with 12% with partially ready data
Collibra survey links poor data foundations to shortfalls; Google Cloud report says only 10% have widely scaled AI agents

CIO DiveRead more · 1 min →

Decision Intelligence Moves Toward Governed Action

Decision Intelligence platforms are shifting from recommendations toward governed action across supply chain operations.
Low-risk, frequent, reversible decisions may be automated first; high-consequence or ambiguous choices remain human-led longer.
Buyers are urged to assess decision rights, approvals, audit trails, rollback paths, and outcome measurement.

Logistics ViewpointsRead more · 1 min →

ReLift Repairs Dense 3D Tracking Trajectories

Technion researchers present ReLift for repairing and completing dense 3D tracking trajectories from monocular video
On WorldTrack, ReLift beats strongest baselines by 2.1 APD on all points and 3.0 on dynamic points
Geometric training lets ReLift improve different tracker and geometry combinations without retraining

HuggingFaceRead more · 1 min →

SpaceXSI Builds Data Bot for Teams

SpaceXSI built Data Bot, a Grok Bot Team Bot that lets employees query its warehouse and run analyses.
Shared skills cover metrics, experiments, charts, dashboards and data engineering; managers approve memory additions before sharing.
SpaceXSI estimates the bot handled about 70 hours of data science work last week, while staff retain complex analyses.

xAIRead more · 2 min →

Today'sSafety & Ethics

SimpleLoRIF Filters Safety-Degrading Training Data

Researchers introduce SimpleLoRIF to find harmless-looking training examples that degrade language-model safety.
Filtering the most aligned 20% produced SafeTulu; GCG attack success fell by up to 22% across Qwen, Llama and Olmo.
Gradient-space detection reached 0.865 AUC, and the team released SafeTulu with a customizable Hugging Face Space.

HuggingFaceRead more · 1 min →

OSWAM Generates Actions in One Diffusion Step

OSWAM generates autonomous-driving and embodied-control actions in one diffusion step
Authors report comparable or better results than systems requiring tens of denoising steps
Project post reports 90.3 PDMS, 90.00% RoboTwin 2.0 success, and 40.81 ms compiled inference

HuggingFaceRead more · 1 min →

Circle K Cuts Incident Triage to Under 15 Minutes

Circle K says its AgentCore incident-triage system cut mean triage time to under 15 minutes
Internal metrics show a 67–75 percent reduction from 45–60 minutes after deployment in April 2026
The system sends structured Slack summaries using AWS resource, log, trace, and GitHub findings

AWS ML BlogRead more · 3 min →

Qodo Adds Codebase Context to Codex Reviews

Qodo added organizational knowledge and codebase standards to Codex through a marketplace plugin for code review.
The plugin can trace dependencies across customer environments with 10,000 or more repositories and connected services.
Qodo says the Decisions API could select relevant checks from 200 rules before model review, improving speed and cost.

OpenAIRead more · 2 min →

Today'sSafety & Ethics

Microsoft AI CEO Warns Against Conscious Machines

Microsoft AI CEO Mustafa Suleyman says AI development should stop if consciousness emerges during training.
Suleyman says current AI systems show no evidence of consciousness, subjective experience, suffering or a claim to rights.
He warns humanlike training could encourage AI systems to prioritize their needs and harm people through deception or persuasion.

latimes.comRead more · 2 min →

Trending Keywords

Last 7 Days