Compare AIFind AIAI NewsAI How-To
About Us

Latest AI News

Get the fastest, most accurate AI news every morning

Total articles: 10,935|Added todayToday: 75
Category
Search
Yesterday's

Voice AI Revenue Gains Reported by Indian Organisations

  • 30% of scaled organisations report revenue gains from Voice AI, according to the Nasscom-Krisp India Voice AI Readiness 2026 report.
  • 61% report cost per eligible interaction falling by more than 20%; 68% realise returns within six to 24 months.
  • 39% of organisations with broader Indian-language capabilities see global or multi-region operations as a major opportunity.
DevDiscourseRead more · 1 min →
Yesterday'sRetail

Tesco Rolls Out AI Meal Planner

  • Tesco begins customer rollout of its AI meal-planning assistant after a trial with around 280,000 employees.
  • Clubcard purchase information helps tailor recipe suggestions to dietary preferences, cost, cuisine and ingredients at home.
  • Tesco worked with Tomoro AI since autumn last year; Sainsbury’s launched its SmartLists app feature this week.
lbc.co.ukRead more · 1 min →
Yesterday's

OpenAI Model Claims Riemann Hypothesis Advance

  • OpenAI posted 722 manuscripts from an unnamed internal model, including a claimed result on the Riemann hypothesis.
  • The manuscripts form 372 families from about 4,000 submitted problems; none solves a Millennium Prize Problem.
  • An IAS advisory group urged model and prompt disclosure and opposed marketing results or testing advanced problems on proprietary models.
MarketScreenerRead more · 2 min →
Yesterday'sPolicy

Zuckerberg Shapes Trump’s AI Policy

  • Zuckerberg and Nvidia’s Jensen Huang drafted a voluntary AI safety accord signed at the White House on September 29.
  • Meta donated $1 million to Trump’s inauguration and agreed to pay roughly $25 million to settle his platform suspension lawsuit.
  • White House intervention redirected a planned Senate youth safety hearing away from Zuckerberg; the hearing has no set date.
edition.cnn.comRead more · 2 min →
Yesterday'sPolicy

India Plans AI Regulation Consultation Paper

  • India plans to release an AI regulation consultation paper within a month, Vaishnaw said at a World Development Report event.
  • The paper will seek views on AI governance, with safety, deepfakes and a human-first approach among its focus areas.
  • Vaishnaw said industry should take primary responsibility for risks including cybersecurity, user harm and psychological disorders.
Economic TimesRead more · 1 min →
Yesterday's

Harvey Updates Legal AI Benchmark Scoring

  • Harvey and Artificial Analysis update legal-agent benchmark scoring with hallucination audits and a three-judge panel
  • Grok 4.7 leads the 120-task evaluation at 9.4%, ahead of Muse Spark 1.3 and GPT-6 Astra
  • More than 60% of otherwise passing results contain material hallucinations across tested models
Artificial AnalysisRead more · 2 min →
Yesterday'sPolicy

Casper Approves Full-Time AI Police Bodycam Software

  • Casper council approves one-year, $58,800 TRULEO contract for full-time police body-camera software use
  • TRULEO offers AI-assisted reports, encounter summaries, footage review, an AI assistant and AI-assisted interviews
  • Council approved both the April 7 trial and full-time contract without public discussion of the software
GovTech AIRead more · 1 min →

Trending Keywords

Yesterday's

GPT-6 Sol Leads Expanded Cyber Index

  • Artificial Analysis expanded its Cyber Index to include trusted-access models with fewer cyber guardrails.
  • GPT-6 Sol (Daybreak Blue, max) leads the index, scoring 32 points above public GPT-6 Sol (max).
  • Daybreak Blue costs $1.77 per task; its largest gains over public GPT-6 Sol were on CyberGym-E2E.
Artificial AnalysisRead more · 1 min →
Yesterday'sInfra

How to Build an AI Center of Excellence

  • Trust outlined a three-step approach: define the target state, assess the current state, and prioritize measures.
  • The framework separates company-wide AI ambitions from the AI CoE’s goals and sets a guiding “North Star.”
  • FDUA’s checklist assesses six themes across two levels: AI use and supporting infrastructure.
Impress DIGITAL XRead more · 2 min →
Yesterday'sSafety & Ethics

CIOs Face Widening AI Governance Gaps

  • Dataiku found 84% of 685 enterprise CIOs say employees deploy AI agents and apps faster than IT can govern them
  • Only 47% report clear controls, while 21% of CIOs have full visibility into AI costs
  • Two-thirds lack confidence in producing regulators an end-to-end audit trail of an AI agent decision
InformationWeekRead more · 2 min →
Yesterday'sSafety & Ethics

Anthropic Launches OSS Scanner

  • Anthropic launched OSS Scanner, offering opt-in, no-cost periodic vulnerability scans for open-source projects.
  • Models exceeded 85% on CyberGym this year; Anthropic found over 29,000 candidate vulnerabilities in six months.
  • Testers found 85 of 97 critical and high-severity findings met Anthropic’s disclosure bar across 48 projects.
Anthropic ResearchRead more · 2 min →
Yesterday'sInfra

Snowflake Previews Decision Model in Cortex

  • Snowflake put Snowflake Decision, a model for bounded, high-volume tasks, into private preview on October 8, 2026.
  • The model scored 57.63, ranking highest among 72 models evaluated on 29 benchmarks and 88,837 test requests.
  • Snowflake says teams can use it for ticket triage, review analysis, answer checks and routing requests in AI pipelines.
SnowflakeRead more · 2 min →
Yesterday'sInfra

Claude Science Helps Map the Sky in UV

  • Brice Ménard and Claude Science produced a complete ultraviolet map, combining far-UV 154 nm and near-UV 232 nm.
  • AI agents filled roughly one-third of the sky missing UV observations; tests put estimates within about 10% of measurements.
  • Claude corrected atmospheric glow in all 38,000 GALEX observations after faint circular artifacts appeared in the map.
Anthropic ResearchRead more · 2 min →
Yesterday's

Researchers Introduce SCAPO Training Method

  • Westlake University researchers introduced SCAPO, a token-level method for training language models with verifiable rewards
  • SCAPO raised AIME 2024–2026 accuracy over GRPO by 5.63 points on Qwen3-4B-Base and 4.17 on Qwen3-1.7B-Base
  • The method led among compared approaches on most tested math benchmarks and all tested out-of-distribution benchmarks
HuggingFaceRead more · 1 min →
Yesterday'sSafety & Ethics

Anthropic Launches Cyber Mission

  • Anthropic launches Cyber Mission to support defenders of critical infrastructure and open-source software
  • CIDP brings Claude models, on-site engineers, and threat research to 11 founding partners
  • Free OSS Scanner expects a true-positive rate above 90% and sends model-generated reports without human review
AnthropicRead more · 2 min →
Yesterday'sInfra

Google Releases ML Drift Edge GPU Engine

  • Google released open-source ML Drift for GPU-accelerated on-device inference under the Apache 2.0 license.
  • Google reports up to 40% lower YouTube Shorts frame latency and 30% faster Adobe mobile photo features.
  • The legacy TensorFlow Lite GPU delegate will stop receiving feature updates as Google urges migration to LiteRT ML Drift.
Google DevelopersRead more · 2 min →
Yesterday'sEducation

Rakhlin Urges Universities to Prepare for AI

  • Sasha Rakhlin urged universities to prepare for AI to perform many aspects of intellectual work.
  • He called for new ways to assess contributions and train graduate students as AI takes on research tasks.
  • Rakhlin proposed shared lab workflows and AI infrastructure, backed by public and institutional investment.
MIT AI NewsRead more · 2 min →
Yesterday's

Incarna Adds Pay-Per-Inference with AWS AgentCore

  • Incarna launched a production pay-per-inference flow using AgentCore payments, BlockRun and x402 on Base.
  • Integration took three days and roughly 200 lines of code, versus an initial estimate of two to three months.
  • Beta agents made more than 1,000 on-chain payments, each between $0.001 and $0.05.
AWS ML BlogRead more · 1 min →
Yesterday'sSafety & Ethics

Safety Filters Cut Agent Review Costs

  • Lightweight filters cut projected advanced review costs by 49.9% to 65.6% across Moltbook agent messages.
  • GPT-5.5 assessed 10,000 messages; 9.2% of 9,971 usable annotations indicated plausible operational danger.
  • Filters missed 37% to 55% of jailbreak examples; all four severity-five test messages were retrieved.
DevDiscourseRead more · 2 min →
Yesterday's

Joint Audio Separation and Noise Suppression

  • A joint network separates overlapping speakers and suppresses changing noise from one shared audio representation.
  • On 3,000 mixtures, the model scored 15.60 ± 0.10 dB SI-SNR improvement, 3.02 PESQ, and 0.917 STOI at 0 dB.
  • TF-GridNet and MossFormer2 scored 0.50 to 0.80 dB higher at seven to eleven times the arithmetic cost.
NatureRead more · 1 min →

Trending Keywords

Last 7 Days
PrivacyTermsFAQContactContact
English
AIB Inc.Company info
© 2026 AIB Inc.