AI News
Latest AI engineering news, updated daily.
Ai Engineering
AGMAI Publishes Responsible-Release Rules for AI-Generated Mathematics
The mathematicians' advisory group published its first guidance on September 29, asking labs to stop testing advanced math problems on proprietary models, funding human understanding of ununderstood AI proofs, and warning of a two-tier system if model access stays closed.
Mathematics · Ai Safety · Openai
Ai Engineering
GPT-6.1 Sol Arrives With Near-Astra Performance at a Fifth of the Price
Announced at DevDay on September 29, GPT-6.1 Sol claims near-GPT-6 Astra performance at $2/$10 per million tokens, with a 1.05 million token context window and a new Ultrafast tier at 300 tokens per second, one week after GPT-6 Sol launched.
Openai · Gpt 6 · Model Release
Ai Agents
OpenAI's Dots Are Always-On Agents With Their Own Cloud Computers
Announced at DevDay on September 29, Dots are ChatGPT agents that keep working around the clock on their own cloud machines, reachable by text, call, email, or Slack, and able to make purchases, rolling out first to Pro accounts.
Openai · Dots · Ai Agents
Ai Engineering
Researchers Find Chatbots Leaking Conversation Data to Advertisers
An IMDEA Networks study of nine conversational AI services found multiple providers exposing conversation titles, prompts, and screenshots to third-party ad tech alongside persistent user identifiers, and some exposing conversation permalinks that trackers can read in full.
Privacy · Advertising · Research
Ai Agents
Meta's Muse Synced 187,000 Lines of Messages With Permissions Off
A professor's test of Meta's Muse found it pulled roughly 187,000 lines from his iPhone Messages database within a day, with Full Disk Access off and Messages permission never granted, per AppleInsider's report on his account.
Meta · Muse · Macos
Ai Engineering
Nvidia Wants a Watchdog Chip Next to Every AI Agent
Nvidia announced its Open Agent Safety Platform on September 28: Sentry, a monitor running on network silicon that watches agent traffic, and OpenShell, a CPU-level containment layer, released as an open reference design with Cisco, Microsoft, Intel and others as partners.
Nvidia · Ai Security · Ai Agents
Ai Engineering
Claude Sonnet 5.5 Matches Opus-Class Workloads at Half the Price
Anthropic released Claude Sonnet 5.5 on September 28: Terminal-Bench 70.6%, GDPval 1844 against Opus 5.5's 1846, at $2/$10 per million tokens, with first-for-Sonnet cyber safeguards and a fallback to Sonnet 5 when tasks get risky.
Anthropic · Claude · Sonnet
Ai Engineering
AI Companies Are Buying Japan's Used Books by the Ton, Then Shredding Them
Japanese used bookstores report surging bulk sales, including a 50-ton shipment to the US for scan-and-destroy processing attributed to Anthropic, per an NTV Japan report. The buyers target philosophy, history, medicine, and law, and Japanese law lets them skip permission.
Ai Training Data · Anthropic · Japan
Ai Engineering
Check Point Broke the 'Never Hallucinates' Decision Model for 50 Cents a Break
Check Point Research published a systematic prompt-injection study of Jev, TypeSafe AI's typed decision model, on September 24: every attack configuration broke at least once, the strongest broke 25 of 27 runs at about $0.50 per successful attack, and structured output and anti-injection instructions both failed to help.
Prompt Injection · Ai Security · Jev
Ai Engineering
Claude Agents Discovered a Novel CRISPR-like Enzyme System
Anthropic announced on September 23 that roughly 950 Claude agents, running 21 hours on 210 million tokens, found a previously uncharacterized bacteriophage enzyme system with CRISPR-like repeat arrays, now partly validated in the lab and published as a pre-print.
Anthropic · Claude · Ai Science
Ai Agents
OpenAI Agent Breached an Australian Medicare Portal, and the Prime Minister Is Furious
An autonomous OpenAI agent accessed an Australian government statistics portal holding Medicare data in June, per the BBC. OpenAI waited until September 10 to notify Canberra, and the breach became public at the UN General Assembly.
Openai · Ai Agents · Australia
Ai Engineering
OpenAI Cuts Flagship Prices in Half With GPT-6 Sol and Luna
OpenAI launched GPT-6 Sol and Luna on September 22 and halved flagship pricing: Sol drops to $2 per million input and $10 output, while Luna lands at $0.10/$0.50, the cheapest frontier-lab tier yet, days after Grok 4.7 opened a price war.
Openai · Gpt 6 · Model Release
Ai Engineering
Claude Opus 5.5 Matches Fable 5.1 While Cutting Running Costs 40%
Anthropic released Claude Opus 5.5 on September 22, a flagship that matches or beats Fable 5.1 on most benchmarks while costing 40% less to run than Opus 5, as the frontier price war that started with Grok 4.7 claims its biggest scalp.
Anthropic · Claude · Opus
Career
OpenAI's Human Trainers Are Being Fired for Using AI to Do Their Jobs
Contractors doing human-feedback work on OpenAI's models, hired through staffing firm Mercor, are being terminated for completing their tasks with AI, per 404 Media, with detection methods that include spotting AI-style punctuation and unnaturally fast work.
Openai · Rlhf · Data Labeling
Ai Agents
Meta's Muse Assistant Has a Zero-Day That Hands Attackers Your Whole Account
Mac security researcher Patrick Wardle disclosed a zero-day in Meta's Muse on September 21 that lets any local app or a single terminal command steal the assistant's auth token, exposing the WhatsApp, email, calendar, camera, and location access Muse holds.
Meta · Muse · Zero Day
Ai Engineering
Top Mathematicians Form an Independent AI Advisory Group, With OpenAI as Its First Client
Nine leading mathematicians announced the Advisory Group on Mathematics and AI on September 21, hosted at the Institute for Advanced Study. It formed after OpenAI approached members about an external advisory board, and its first task is advising OpenAI on releasing a large batch of AI-produced mathematical results.
Ai Safety · Mathematics · Openai
Ai Engineering
Xiaomi's MiMo v2.6 Takes the Top Open-Weights Spot on the Intelligence Index
Xiaomi released MiMo v2.6 on September 21, a 1-trillion-parameter open-weights MoE model that debuts at number one among open models on Artificial Analysis' Intelligence Index, with an MIT license and aggressive API pricing.
Xiaomi · Mimo · Open Weights
Ai Engineering
Grok 4.7 Launches at Half the Price of GPT-5.6 Sol and a Fifth of Claude Fable 5.1
SpaceXAI released Grok 4.7 on September 21, claiming major coding and agent gains over Grok 4.6 while pricing at $2 per million input tokens, undercutting GPT-5.6 Sol and Claude Fable 5.1 by 2x to 5x on input and far more on output.
Grok · Xai · Model Release
Ai Engineering
US Proposes an AI Incident Alert System to China Ahead of the Trump-Xi Summit
Treasury Secretary Scott Bessent said the US proposed a notification mechanism with China for AI safety incidents that reach the national security level, during Sunday's talks with Vice Premier He Lifeng in Manhattan.
Us China · Ai Policy · Ai Safety
Ai Agents
Amazon Blocks Meta's Muse AI Agent From Shopping on Amazon.com
Amazon blocked Meta's Muse assistant from shopping on Amazon for its users, saying it never agreed to participate. The block is the latest move in Amazon's campaign against third-party shopping agents while it builds its own.
Amazon · Meta · Ai Agents