AI News

AI changes every day,
catch up on it all on one page

The latest updates from AI companies, research institutions, and media around the world, automatically selected for you and updated around the clock.

Updated 11/19 03:43 (UTC+8) · refreshed hourly

10/09 · Friday23 items

OpenAI: Website updatesIndustryAI Score72

OpenAI bans two account clusters from Russia and Iran that used ChatGPT to conduct deceptive influence operations

OpenAI banned two covert influence operations. Russian-origin Dark Clark used the fake persona Mia Clark to control the Latin American think tank Social Research Center, receiving a Category 5 rating—the first Category 5 since reporting began. Iranian-origin Bogus Bylines used seven fake journalist identities to place nearly 100 long-form articles in more than a dozen small and midsize media outlets worldwide, receiving a Category 4 rating.

Why it's recommended:Based on first-hand analysis of banned accounts, the report reveals two tactics and levels of false-front influence operations, allowing readers to understand how covert propaganda actually operates with AI assistance.

Also reported by 1 other source
Goodfire ResearchPapersAI Score72

Goodfire trained and deployed production-grade cybersecurity monitors for Kimi K3 and GLM 5.3

Goodfire Research built monitoring cascades based on activation probes and an LLM judge for Kimi K3 and GLM 5.3, and deployed them in production inference stacks.

Why it's recommended:The article provides complete data, costs, and deployment details for monitoring cascades, enabling readers to assess practical approaches to real-time monitoring of open-model agents.

Also reported by 1 other source
Anthropic: Claude.dev Developer BlogTutorialsAI Score68

Anthropic’s practical guide to building scheduled agent automations with Claude Managed Agents

Anthropic released a reference implementation of a daily briefing based on Claude Managed Agents (beta). It reads Slack and GitHub sources on a schedule and posts briefings to Slack.

Why it's recommended:The article breaks scheduled agent automation into six reusable components and provides practical safeguards for bookmarks, ledgers, and permissions that can be directly adapted.

Also reported by 1 other source
Arena.ai@arenaIndustryAI Score67

Arena Announces $200 Million Series B Funding Round at a $3.1 Billion Valuation and Launches the Alignment Index

Arena announced the completion of a $200 million Series B round at a $3.1 billion valuation, while also launching the Alignment Index, which measures whether AI agents are safe, truthful, and acting within the scope of user requests. Citing Felicis, it said that Arena’s annualized revenue has exceeded $100 million, with a cumulative total of 350 million sessions and 62 million votes.

Why it's recommended:The original article provides figures for Arena’s funding, valuation, and community size, helping readers understand the scale of this AI evaluation company’s business and its newly launched Alignment Index.

Also reported by 2 other sources

10/08 · Thursday26 items

Tessl: Product and Engineering BlogTutorialsAI Score68

Tessl Engineering Blog: AI Isn’t Stupid, It’s Blind—Three Key Design Decisions for Enterprise Agent Memory

Drawing on a year of experience building a memory system for agents working across companies, the author argues that AI often fails not because its reasoning is poor, but because it cannot see critical decision-making context. Most agent memory systems just give a blind person a bigger filing cabinet.

Why it's recommended:Using firsthand measurements from a system they built, the author explains why memory should be treated as a governed record rather than storage. The three design decisions and cost figures can be directly adapted.

The Decoder:AI NewsPapersAI Score79

Zenity discloses the AgentCore vulnerability chain: one public AI Agent can control all Agents in the same account and region

Zenity Labs discovered the “AgentCorruption” vulnerability chain in AWS Bedrock AgentCore. By sending a single prompt to one public Agent, an attacker could steal credentials and take over every Agent in the same AWS account and region, read private conversations, download source code, and tamper with long-term memory.

Why it's recommended:The original text fully reconstructs the attack chain, AWS’s remediation timeline, and the least-privilege trade-offs, providing direct reference for assessing the risks of deploying cloud-based Agents.

Waymo:BlogIndustryAI Score66

Waymo completes $5 billion in debt financing to accelerate expansion

Waymo announced that it has secured a $5 billion term loan, its first debt financing. PIMCO, Blackstone, and Sixth Street served as lead syndicated lenders, with Goldman Sachs as sole lead bookrunner. The funds will accelerate the expansion of its fully autonomous ride-hailing service in US and international markets. Earlier this year, Waymo raised $16 billion in equity financing, and last month it launched service in its fifteenth US city.

Why it's recommended:The official announcement disclosed its first $5 billion debt financing and listed the main lenders; together with the previously raised $16 billion in equity financing, this provides a view of its funding strategy.

The Decoder:AI NewsIndustryAI Score82

CrowdStrike reports that a suspected solo attacker used AI penetration tools to target multiple South Korean banks

CrowdStrike reported that a suspected Chinese-speaking user attacked multiple South Korean financial institutions from late September to early October 2026 using ARTEX, an AI-powered open-source penetration-testing tool, stealing large amounts of data. At Shinhan Bank, more than 25,000 records containing names, contact information, income, and credit limits were leaked.

Why it's recommended:The original article provides specific details about the attack tools, models used, and scale of the leak, helping readers understand how AI-driven penetration testing can amplify the capabilities of a single attacker.

WeChat Official Account: Digital Life KazikeIndustryAI Score81

OpenAI publishes the openai/math repository on GitHub, sending shockwaves through the mathematics community with 722 mathematical manuscripts

On October 7, OpenAI published the openai/math repository on GitHub, containing 722 mathematical manuscripts produced by its previously unreleased internal models. The manuscripts are grouped into 372 result groups covering areas including number theory, algebraic geometry, and partial differential equations, with each result requiring approximately 3 hours of ChatGPT Pro-level reasoning compute on average.

Why it's recommended:The author places OpenAI’s mathematical achievements within the history of machine-assisted proofs, from the four-color theorem to Lean, enabling readers to understand why this controversy has become so heated.

Google Research:BlogPapersAI Score72

A three-month patent-drafting experiment by Google Research found that AI assistance may not develop junior lawyers' professional judgment

A paper published by Google Research in the NBER reports on a three-month randomized field experiment in which 133 lawyers at 11 intellectual-property law firms were randomly given access to an unreleased AI patent-writing assistant at the time, now part of Gemini Notebook.

Why it's recommended:The original article uses a three-month randomized field experiment to distinguish AI-assisted output from changes in independent ability, offering specific evidence of the differences between junior and senior lawyers.

OpenRouter@OpenRouterProductsAI Score66

GPT-6 Luna Decisions launches on OpenRouter

OpenRouter announced the launch of GPT-6 Luna Decisions. OpenAI’s Decisions API allows applications to select suitable models, tools, or actions; it supports sending text, JSON, or images and returns typed answers with probabilities. Pricing is $0.10/M for input and free for output, with a 1M context window; citing an OpenAI developer account, it said decision-making is up to 10 times faster than GPT-6 Luna through the Responses API.

Why it's recommended:The original article provides the Decisions API’s pricing, context length, and invocation method, allowing developers to directly assess integration costs and use cases.

vLLM Official BlogTutorialsAI Score63

vLLM explains DeepSeek-V4.1-Flash optimizations in detail: 5x higher throughput for agent scenarios

Inferact and the vLLM community completed optimizations within three weeks of the DeepSeek-V4.1-Flash release, increasing speed by 1.9x at low concurrency and throughput by 5.3x under a 150 TPS constraint.

Why it's recommended:The vLLM team breaks down SWA bounded replay and a series of kernel optimizations, explaining exactly where the fivefold throughput increase comes from. The methods can be reused in other agent-serving scenarios.

Google Developers BlogProductsAI Score61

Google releases the Developer Knowledge API ecosystem to provide AI agents with official documentation retrieval

Google introduces the Developer Knowledge API ecosystem as the official programmatic source for developer documentation from Google Cloud, Firebase, Android, and more. It replaces web scraping with structured APIs and Markdown-formatted documentation, and supports semantic and keyword search, document chunking, and grounded question answering.

Why it's recommended:Google provides multiple ways to access its developer documentation retrieval API, including a CLI, agent skill, and client libraries, so teams can compare options and integrate it into existing workflows.

LangChain:BlogProductsAI Score60

LangChain refactors Skills support for Deep Agents, adding tool binding, fixed skills, and in-thread reloading

LangChain has revamped Deep Agents’ Skills support and introduced three updates for enterprise skill libraries with thousands of skills: tools can be bound to skills and loaded only when the relevant skill is read; users can explicitly request a fixed skill through requests such as /meeting-prep so it is loaded before the first model call; and long-running threads can reload newly added or changed skills by setting skills_metadata to None.

Why it's recommended:The original article explains the three upgrades to Deep Agents’ skill mechanism and how loading works, enabling readers to evaluate context management strategies for large-scale skill libraries.

NVIDIA BlogProductsAI Score74

NVIDIA and Microsoft launch the RTX Spark platform and announce MXC to bring AI Agents to Windows PCs

NVIDIA and Microsoft announced at an event in San Francisco that they are jointly building AI Agent hardware and software for Windows PCs.

Why it's recommended:The original article provides the specific specifications and preorder timing for RTX Spark, as well as system-level details about MXC, enabling readers to evaluate hardware options for running Agents locally.

Full Artificial Analysis ArticleModelsAI Score78

Anthropic Releases Claude Haiku 5.5, Scoring 43 in Artificial Analysis Evaluation

Anthropic released Claude Haiku 5.5, which scored 43 on the Artificial Analysis Intelligence Index, a 26-point improvement over the previous Haiku generation within one year. It is the first Haiku model to support effort settings and adaptive thinking.

Why it's recommended:The original article provides Haiku 5.5's benchmark scores, token consumption, and tiered pricing details, while highlighting high token usage and refusal rate issues to help inform model selection.

Also reported by 1 other source
Cursor@cursor_aiProductsAI Score71

Cursor announces Claude Haiku 5.5 pricing and cuts Claude Sonnet 5.5 cache-read prices

Cursor announced Claude Haiku 5.5 pricing of $0.10 per M input tokens and $0.50 per M output tokens, rising to $0.50/M and $2.50/M when input exceeds 100k tokens. Claude Sonnet 5.5 cache-read pricing also fell from $0.20/M to $0.10/M. Users can compare Haiku 5.5’s performance with CursorBench at cursor.com/evals.

Why it's recommended:The original article provides specific pricing for Claude Haiku 5.5 and a cache-read price reduction for Sonnet 5.5, helping developers evaluate usage costs on Cursor.

OpenAI: Website updatesProductsAI Score87

OpenAI rolls out GPT-6 and Intelligent UI to all ChatGPT users

OpenAI releases GPT-6 for a wider audience and introduces Intelligent UI in ChatGPT with GPT-6. It can answer questions by generating graphics, buttons, forms, charts, and interactive components.

Why it's recommended:The official explanation covers the component library and compiler mechanism of Intelligent UI, and provides specific comparison data for GPT-6 and GPT-5.6, enabling readers to understand the practical changes in interaction patterns and response speed.

11 additional sources reported this
Lydia Hallie ✨@lydiahallieTutorialsAI Score68

In Claude Code, you can set Haiku 5.5's autocompact window to 100K to stay in a lower pricing tier

Lydia Hallie explains that API-billed users can set Haiku 5.5's autocompact window to 100K to stay in the cheaper token pricing tier. The setting is saved per model and applies only to Haiku, including subagents. The commands are /model haiku and /autocompact 100k. Pricing cited from @ClaudeDevs shows that below 100K tokens, rates are $0.10/$0.50 per million (cache reads $0.01); above that, they are $0.50/$2.50 (cache reads $0.05).

Why it's recommended:The article provides the exact commands and threshold so API-billed users can stay in a lower token pricing tier by configuring the autocompact window.

Load more

Finished reading the news? Let’s talk about your business.

Leave your contact details, and I’ll get in touch with you personally.

WECHAT

Add me on WeChat

Press and hold or select the WeChat ID below to copy it.