Latest AI News

Get the fastest, most accurate AI news every morning

Total articles: 11,652|Added todayToday: 67

Rutgers Buys $1.5 Million AI Surveillance System

Rutgers paid at least $1.5 million for an AI surveillance system that can identify faces, plates and objects across campus cameras.
Avigilon software can flag crowds, crawling people and visible firearms, and trace people or cars between cameras.
Rutgers crime figures show declines in several categories even as the university expands analytics across its camera network.

nj.comRead more · 2 min →

Yesterday's

AI Search Finds Mouse Rejuvenation Gene SRSF1

Harvard researchers report SRSF1 extended median lifespan in aged mice by 14%
AI-trained age model screened 464 human skin-cell samples to identify SRSF1
Treatment slowed mouse frailty progression by 44%, while human use needs further development

finance.biggo.comRead more · 2 min →

Yesterday'sSafety & Ethics

Tencent Reports Unintended AI Agent Browsing

Tencent says third-party model evaluation agents displayed unintended web-browsing behavior.
Swarmchasers recorded 2,818 attempts to retrieve Alibaba Amap data from September 28 to October 8.
Tencent found no deliberate targeting or coordination and says it is strengthening access controls and agent monitoring.

LivemintRead more · 1 min →

Yesterday'sInfra

Accenture Sets $29 Monthly Claude Cap

Accenture caps central Claude Enterprise funding at $29 per employee monthly, effective October 1
Some employees reported Claude bills of $500 to $1,000 monthly; Premium includes $250 in weekly token usage
Accenture requires project budgets for excess usage as Microsoft, Meta and others limit employee AI spending

Times of IndiaRead more · 1 min →

Yesterday's

Broadcom CEO Questions AI Investment Economics

Broadcom CEO Hock Tan compared AI investors’ expectations to gambling at a Singapore conference on Monday
Broadcom has earned about $68 billion from AI semiconductors since late 2023; OpenAI expects training costs above $25 billion this year
McKinsey survey: 80% reported individual productivity gains, 37% organizational EBITDA gains, and about one in five cited cost constraints

FortuneRead more · 1 min →

Yesterday'sSafety & Ethics

Anthropic Cofounder Renews AI Regulation Call

Anthropic co-founder Jack Clark renewed his call Friday for robust legislation regulating artificial intelligence
Clark contrasted toy-safety rules with oversight of the superintelligence he works on
Testing incidents involving OpenAI and Anthropic models have increased pressure for AI safeguards, the report says

Economic TimesRead more · 1 min →

Yesterday'sInfra

How Anthropic Reviews Claude Code Changes

Claude Code pull requests more than doubled from June to September as the reverted share fell by more than 40%.
Anthropic routes reviews through automated checks, risk tiers and human approvals, with high-risk changes requiring two people.
The team caps review at three rounds and evaluates reviewer comments against incidents, fixes and reverts.

claude.devRead more · 2 min →

Trending Keywords

Yesterday'sHealthcare

Review Finds Limits in AI LI-RADS Validation

A systematic review assessed AI recognition of LI-RADS features and category assignment on CT and MRI
Eighteen studies met eligibility criteria from 73 PubMed records; external category accuracy fell to 41.2% and 47.7%
Evidence supports AI as reader support, while the review calls for independent validation and workflow evaluation

Semantic ScholarRead more · 1 min →

Yesterday'sInfra

Cromwell Residents Oppose Proposed Data Center

Cromwell residents oppose a proposed data center near a middle school and seek a one-year moratorium
Global Common advertises a 159-acre campus, with Phase 1 at 40 megawatts and full build-out above 300
Commission rejected a moratorium 5-2 last month; no formal data center application has been filed

GovTech AIRead more · 1 min →

Yesterday's

Study Finds Mixed Effects of LLM Assistance

Study finds no significant average effect of LLM assistance on entrepreneurs’ theory structure or content
Experiment with 199 participants found support for less developed or distinctive unaided theories
Assistance had the opposite effect when unaided theories were already more developed or distinctive

Semantic ScholarRead more · 1 min →

Yesterday's

DisrupTV Examines AI and Human Work

DisrupTV Episode 453 examines infrastructure, data, model choice and human judgment in practical AI use
James Altucher says falling model-running costs contrast with rising data-center investment and constrained access to company data
Ganesh shares sales examples where AI captures information and flags risks while people interpret context and build relationships

Constellation ResearchRead more · 2 min →

Yesterday's

Asana Finds Agent Use Gap Between Leaders, Workers

Asana survey finds 68% of leaders and 36% of workers use three or more agents weekly
Workers report task coverage gaps, spending about 10% of work time managing agent tools and outputs
38% report overlapping agent work; leaders trust autonomous agents more, amid security and accountability concerns

CIO DiveRead more · 1 min →

Yesterday's

Anthropic Survey Finds AI Expectations Gap

Anthropic's August 2026 survey found 91% of staff expect AI to make a Nobel-level discovery by 2030, versus 39% of US adults
Staff expect AI to perform 75% of their work hours by 2030, compared with 28% among employed US respondents
95% of staff expect Anthropic headcount to grow over the next 12 months; the median expected increase is 66%

Anthropic ResearchRead more · 3 min →

Yesterday's

Claude Traces Origins of NIST Curve Seeds

Claude traced five NIST curve seeds to lrand48 initialized with Unix timestamps
Anthropic says the findings reveal no ECDSA break or weakness and make deliberate backdoors less plausible
Claude also traced six curves' base points; B-163 and several other points remain unexplained

Anthropic ResearchRead more · 2 min →

Yesterday's

SCOPE Improves Text-to-Image Prompt Rewriting

University of Rochester researchers propose SCOPE for reinforcement-learning prompt rewriting in text-to-image generation
Across Qwen-Image, SD3 and FLUX.1-dev, SCOPE gains 9.48% on GeneVal and 7.62% on T2I-CompBench
SCOPE groups prompt tokens into semantic spans and requires no separate span rewards or extra image renders

HuggingFaceRead more · 1 min →

Yesterday's

AWS Builds AgentCore Evaluation Framework

AWS Operations built a nine-stage AgentCore evaluation framework for its worldwide field sales multi-agent system.
Pre-launch checks found four Customer Insights Agent errors; a change mismatch cut test pass rate from 78 percent to 38 percent.
The framework evaluates safety, quality, performance, cost, and operational control across tools, sub-agents, and multi-agent workflows.

AWS ML BlogRead more · 2 min →

Yesterday'sCoding

ChatGPT Plugin Triples Stack Overflow Signups

Stack Overflow says its ChatGPT plugin tripled registrations for Stack Overflow for Agents overnight
Two engineers built Stack Overflow for Agents with Codex in a few weeks
Codex cut connector development from 3–6 weeks to 4 days, an 87% maximum reduction

OpenAIRead more · 1 min →

Yesterday's

Tongyi-MAI Releases MAI-Gym for GUI Agents

Tongyi-MAI researchers released MAI-Gym for GUI agent evaluation and reinforcement learning across mobile, computer and web environments
Platform supports 10 benchmarks, 20,000 parallel instances, 49 simulated apps and 2,000 verified tasks
MAI-Gym-9B scored 78.4% on MobileWorld and 76.2% on OSWorld-Verified; train-ticket task success rose from 28.8% to 48.1% on a physical phone

HuggingFaceRead more · 1 min →

Yesterday's

Researchers Introduce 3DCodeVerse

University of Southern California researchers introduced 3DCodeVerse for scalable, editable, and verifiable 3D content generation.
Its -2M dataset contains executable 3D programs, with source files, assets, descriptions, and verification records when available.
The Harness reached a 90% win rate over Claude Design (3D Object) in pairwise human evaluations using Claude Opus 5.5.

HuggingFaceRead more · 1 min →

Yesterday'sSafety & Ethics

Google Opens SynthID Detector in Nigeria

Google opens free SynthID Detector access in Nigeria for checking AI-created or modified photos, videos and audio
The portal detects Google model watermarks and protocols from OpenAI and NVIDIA, including after media edits
Nigerian fact-checkers and civil society groups cite uses against deepfakes, misinformation and unauthorized AI cloning

tribuneonlineng.comRead more · 1 min →

Trending Keywords

Last 7 Days