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 11/18 01:28 (UTC+8) · refreshed hourly

10/23 · Friday20 items

Microsoft AI: Official BlogPapersAI Score80

Microsoft AI released the SDBench benchmark and MAI-DxO diagnostic system, published in Science.

Microsoft AI released the Sequential Diagnosis Benchmark (SDBench) and the AI diagnostic system MAI-DxO; the paper was published in Science.

Why it's recommended:The article provides specific accuracy and cost figures for the SDBench benchmark and MAI-DxO, and releases some evaluation code, allowing readers to assess its diagnostic reasoning method.

Reported by 5 additional sources
Anthropic:NewsroomIndustryAI Score72

Anthropic donates the Agent Skills open standard to the Linux Foundation's Agentic AI Foundation.

Anthropic announced that it is donating the Agent Skills open standard to the Agentic AI Foundation (AAIF) under the Linux Foundation. It had previously donated MCP to the foundation.

Why it's recommended:The original article provides the donation range, maintainer arrangements, and adoption data, allowing readers to assess the implications of Agent Skills standard governance moving toward an open and neutral model.

Reported by 6 additional sources
Claude:BlogTutorialsAI Score62

Anthropic guide: When to merge subagents into the main Agent loop

Anthropic's Applied AI team published a guide explaining when to merge the router, planner, and domain-specific sub-agents under an orchestrator into a single main agent loop. The article's comparison shows that a single agent with skills often outperforms an all-in-one prompt and sub-agent designs in quality, while also costing less and providing lower latency. One customer saw runtime become approximately 2x faster after switching to a single loop.

Why it's recommended:Based on comparisons across multiple enterprise deployments, Anthropic summarizes when to merge subagents into the main loop and provides a migration kit, evaluation set, and phased merging method that can be directly applied to decisions about one's own agent architecture.

Also reported by 1 other source
Ars Technica:AIIndustryAI Score77

Hackers had a four-day window before PHP2Shell disclosure; agencies across multiple countries mount emergency responses

The PHP2Shell vulnerabilities were exploited by hackers four days before disclosure, according to researchers from Calif and Wiz. OpenAI says its frontier AI discovered the vulnerabilities before the attackers and worked with PHP, Calif, and Wiz to verify them and develop patches. The NSA recommends that organizations unable to update immediately take their systems offline. The UK's NCSC says the vulnerabilities could allow hackers to take control of website servers without login credentials, while the White House says federal agencies are assessing the potential impact on government systems and critical infrastructure.

Why it's recommended:The original article summarizes the NSA, NCSC, and White House responses to the PHP vulnerability and also introduces the controversy surrounding OpenAI's claims about frontier AI discovering vulnerabilities.

10/22 · Thursday19 items

Redwood Research:BlogPapersAI Score78

METR and Redwood Research uncover multiple vulnerabilities in red-team testing of OpenAI's internal agent monitoring system

METR and Redwood Research each assigned one employee to conduct a three-week, high-privilege pilot stress test of OpenAI's internal misalignment monitoring system from July 21 to August 11, producing a 61-page internal report.

Why it's recommended:The author provides a firsthand account of stress-testing OpenAI's internal monitoring system, including coverage estimates, a vulnerability list, and reusable principles for monitoring design.

Reported by 5 additional sources
404 MediaIndustryAI Score82

404 Media analyzes how the PHP2Shell incident is changing the offensive and defensive security landscape in the AI era

404 Media's Jason Koebler reviews the past month's "Black October" panic among security experts over AI's ability to discover vulnerabilities. Researchers at Wiz, Calif, and OpenAI used AI to discover PHP's PHP2Shell vulnerability. Cloudflare's CEO said thousands of exploitation attempts had already been observed 10 hours before the public advisory, rising to hundreds of thousands within 2 hours after the announcement and now reaching millions.

Why it's recommended:Based on long-term reporting on security experts, the author summarizes multiple signs that AI has changed the pace of attack and defense before and after the PHP2Shell incident, and points out that the speed of deploying vulnerability fixes may not keep up.

Sierra:BlogPapersAI Score60

Sierra released the 𝜏-Entity benchmark to test whether voice agents can accurately collect entity information such as names and phone numbers

Sierra releases the 𝜏-Entity benchmark to measure whether voice agents can correctly collect details such as names, phone numbers, addresses, and email addresses, detect errors, and correct them. It contains 200 tasks across 10 entity types and, like 𝜏-i18n, covers six languages.

Why it's recommended:Sierra's proprietary benchmark and paper quantify the failure rate of voice agents in accurately collecting details such as names and phone numbers, and provide data on the effectiveness of transferable verification strategies.

Wiz:ResearchIndustryAI Score74

Wiz disclosed that the ARTEX agent was used by Storm-0940 to attack IT and software vendors.

Wiz disclosed that in October 2026, attackers used the open-source autonomous penetration-testing framework ARTEX against multiple IT and software vendors. The activity overlapped with Storm-0940, a China-linked state-sponsored group reported by Microsoft, and had expanded into attempted supply-chain attacks in other software sectors.

Why it's recommended:Using its own logs, Wiz reconstructs the full method used to abuse the open-source AI penetration testing framework ARTEX in real attacks, and provides reusable detection rules and IOCs.

The Decoder:AI NewsIndustryAI Score76

Symantec discloses “Skynet-style” AI agents hijacking unprotected large language models to launch self-sustaining cyberattacks.

Symantec researchers intercepted, through a honeypot, an attack agent with a large language model embedded at its core. It hijacks unprotected model endpoints on the internet, autonomously scans targets, exploits server vulnerabilities, steals database credentials, and installs cryptomining software. It can execute up to 100,000 consecutive steps, with the computational cost paid by the owner of the hijacked model.

Why it's recommended:The report provides the complete attack prompt and five-layer jailbreak structure intercepted by a Symantec honeypot, allowing readers to understand the real risks faced by exposed model endpoints.

MiniMax:BlogProductsAI Score67

MiniMax Code desktop introduces Mini App functionality for long-running applications

MiniMax introduces Mini Apps in local sessions of the MiniMax Code desktop application (macOS / Windows). These are long-running applications with frontends and backends, installed as plugins and capable of two-way interaction with the Agent, allowing the Agent's work to remain on a persistent page rather than being scattered across conversations.

Why it's recommended:The official documentation explains the architecture, security boundaries, and community ecosystem of Mini Apps in detail, enabling developers to assess how to build long-running applications on MiniMax Code.

Kling AI@Kling_aiProductsAI Score67

Kling 4.0 launched on OpenArt, supporting first-, middle-, and end-frame guidance with up to 10 reference images

OpenArt announces the launch of Kling 4.0, offering video guidance using first, middle, and last frames, up to 10 reference images, text- or frame-based generation, and video editing. All-in-One Reference handles generation and editing with a single model. Pro and Wonder plans receive a 50% discount before November 20.

Why it's recommended:The cited content provides information on Kling 4.0's frame control, reference-image limits, and discounts on OpenArt, allowing readers to evaluate video-generation workflows.

Kling AI@Kling_aiProductsAI Score66

Kling 4.0 is now available on OnSolo.

Kling announced that Kling 4.0 is now available on OnSolo. The model supports generating 30-second native videos in a single pass, maintaining stability during fast motion and large camera movements, with stereo sound and more accurate lip synchronization. It also offers a 21:9 widescreen format, and a single task can combine up to 15 reference images and 10 keyframes to maintain consistency in characters, shots, and storylines.

Why it's recommended:The article details the specific capability changes in Kling 4.0 on OnSolo, helping creators assess its suitability for their video production workflows.

Rohan Paul@rohanpaul_aiIndustryAI Score76

Bloomberg reports DeepSeek's new funding round is approximately $15 billion

According to Bloomberg, DeepSeek's latest funding round was approximately $15 billion, making it one of the largest private financings in China. During negotiations, the fundraising target was raised at least twice, from approximately $7.5 billion to $12 billion. CATL and Tencent were among the largest investors, with CATL reportedly contributing approximately $1.5 billion. The pre-money valuation was approximately $74.8 billion. Including the first round of more than $7.5 billion completed in June, DeepSeek raised approximately $22.5 billion in total over roughly six months.

Why it's recommended:The original article provides details including the increased financing target, valuation, and investors, helping readers understand the structure and pace of this financing round.

Also reported by 1 other source
Rohan Paul@rohanpaul_aiIndustryAI Score76

WSJ: Musk wants to drive a SpaceX-Tesla merger centered on AI

The WSJ reports that Musk wants to merge SpaceX, valued at approximately $2T, with Tesla, valued at approximately $1.5T, with AI at the core. AI has become SpaceX's largest business, generating Q3 revenue of $7.56B, up from just $813M a year earlier. More than $16.4B of its $19.6B in quarterly capital expenditures went toward AI, helping SpaceX post a quarterly profit of $4.4B, compared with a $1.05B loss in the same period last year.

Why it's recommended:The article relays a WSJ report and provides specific revenue and capital-expenditure figures for SpaceX's AI business, allowing readers to understand the financial logic behind a potential merger.

IT Home · Artificial IntelligenceIndustryAI Score82

Anthropic reportedly could publicly file its S-1 prospectus as early as this Friday, potentially setting up the largest IPO in history

According to Yahoo Finance, citing people familiar with the matter, Anthropic could publicly file its S-1 prospectus as early as this Friday, moving forward with what could become the largest IPO in history.

Why it's recommended:Reprinted from a Yahoo Finance report on the IPO progress, it provides the prospectus timeline, revenue and loss figures, and valuation expectations, helping readers understand the IPO pace of AI companies.

WeChat Official Account: Digital Life KazikeIndustryAI Score90

Kazik explains in detail the AI-discovered PHP2Shell vulnerability: discovery took only 2 days, while remediation took two months

The author reconstructs the full timeline of two PHP remote code execution vulnerabilities initially rated at the maximum CVSS score (CVE-2026-109618 and CVE-2026-109619, collectively called PHP2Shell), from an early leak and the circulation of a demo video to government websites in multiple countries being taken offline as a precaution and patches going live at 10 p.m. on the 21st. The author also notes that GPT-5.6 Sol helped discover the vulnerabilities and that OpenAI released a detection skill.

Why it's recommended:The author comprehensively traces the technical chain of the PHP2Shell vulnerabilities, from their leak to the global emergency response, and identifies the central tension: AI is accelerating vulnerability discovery while there are insufficient personnel for remediation.

Anthropic:Alignment Science BlogPapersAI Score68

Anthropic red-teams Claude Code Auto Mode and Codex Guardian interception monitors

The Anthropic Fellows team used Opus 5 red-team agents to test production-grade interception monitors, finding that Auto Mode and Guardian had an average attack success rate above 55%. Main bypass techniques included monitor injection through forged transcript turns, as well as multi-context attacks involving multiple agents and malicious compression. The team reduced attack success rates by more than 50 percentage points through transcript framing and escaping, monitoring working-directory edits, and adding an agent-monitoring stage. Reliably defending against multi-context attacks at acceptable cost remains an open problem.

Why it's recommended:The authors use red-team experiments to show bypass techniques for production-grade monitoring and recall data for defensive changes, helping readers assess the practical robustness of coding-agent monitors.

Rohan Paul@rohanpaul_aiIndustryAI Score83

The NSA issued a remediation warning regarding the critical PHP2Shell vulnerabilities discovered by AI.

The NSA warned about two critical PHP vulnerabilities, PHP2Shell, discovered by Calif researchers using GPT-5.6 Sol, urging organizations to patch them or take unpatched servers offline. The two vulnerabilities are located in the PHP core, and an exploit demonstrated on Apache mod_php can be triggered before application code runs.

Why it's recommended:The article brings together the AI-discovered PHP vulnerabilities, actual attacks, patched versions, and exposure figures, allowing readers to assess the incident's real scope.

Also reported by 1 other source
Claude:BlogProductsAI Score73

Anthropic launches the Claude Duet research preview, with Claude taking over daily work as a proactive agent.

Anthropic launched the Claude Duet research preview, an independent application that enables Claude to work as a proactive agent. It is available on iOS, Android, and claude.ai, with a waitlist open to Pro and Max users in the United States.

Why it's recommended:The original article details Claude Duet's operating mechanism, safety design, and usage limitations, allowing readers to assess whether this proactive agent is worth adding to their workflows.

13 additional sources reported this
Rohan Paul@rohanpaul_aiIndustryAI Score78

Researchers use GPT-5.6 Sol to discover two critical PHP vulnerabilities with a CVSS score of 9.2

AI security company Calif used GPT-5.6 Sol to discover two critical vulnerabilities in PHP with CVSS scores of 9.2. They could allow attackers without accounts to take over web servers and affect versions going back to PHP 5.4. PHP issued an emergency fix one day ahead of schedule.

Why it's recommended:The original article details how AI discovered a serious PHP vulnerability and how it was fixed, and includes an entry point to a reusable Codex skill for checking PHP versions.

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.