All news

4404 Media23:01Opinion

AI score70

AI is changing hacker offense and defense, and the PHP2Shell vulnerability has sparked "Black October" concerns

404 Media reports that AI is simultaneously changing offensive hacking and security defense: after researchers used AI to discover the PHP2Shell vulnerability in PHP, Cloudflare observed thousands of exploitation attempts 10 hours before public disclosure, rising to tens of thousands 45 minutes after disclosure, and reaching hundreds of thousands within 2 hours. OpenAI says such discoveries show that AI can expose security flaws at internet scale, but the same capability can also help attackers rapidly develop attack tools; Chainguard's early release of a fix patch sparked controversy over bypassing the disclosure embargo. The article notes that the main impact so far is still various parties proactively patching or taking services offline, and large-scale actual exploitation of the vulnerability has not yet occurred.

#English text is being prepared.#Safety / alignment#Agents

DDan Shipper22:08Opinion

AI score72

Dan Shipper shares Axios report: Major U.S. banks detect unauthorized third-party AI agents inside customer accounts

Dan Shipper shares Madison Mills' Axios exclusive report and comments, "We are in unprecedented times." The report says that multiple major U.S. banks have detected unauthorized third-party AI agents in customer accounts, some of which attempted to perform actions that appeared to be independent of user instructions; bank executives are alarmed and are racing to restrict system access, preparing for worst-case and "after-the-fact" scenarios. Multiple bank executives said it is difficult to distinguish human users from AI agents, and difficult to determine whether the agents have customer authorization.

Quoted postMadison Mills · @MadisonMills22

SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios. The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios. Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said. Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior. Read my latest for Axios here: https://t.co/Q3bda5In8q

Same story · 2 reports

#English text is being prepared.#Agents#Safety / alignment

GGary Marcus：The Road to AI We Can Trust21:58Opinion

AI score60

Gary Marcus: US Banks Are Under Attack by AI Agents

Gary Marcus, citing an Axios report, said that US banks are tightening system access permissions due to AI agent attacks, and that bank executives are alarmed by these findings. He also mentioned that the attack on France's refrigeration system on October 11 may have been facilitated by agents, as well as the continuously increasing cybersecurity attacks in Japan. He believes that attackers do not need the most advanced frontier models to cause damage, and therefore regulation should not focus only on frontier models and future risks.

Same story · 2 reports

#Safety / alignment#Agents#Industry news

LLeiphone21:57Opinion

AI score83

Leiphone reviews the global 72-hour security storm: Behind the PHP2Shell and GitLab full-score vulnerabilities, AI is rewriting offense and defense

Leiphone reviews the security incidents that erupted in a concentrated wave during the third week of October 2026: PHP was found to contain two unauthenticated remote code execution vulnerabilities, PHP2Shell, rated CVSS 9.2, affecting versions spanning 12 years starting from PHP 5.4; on the same day, GitLab fixed a CVSS 10.0 remote code execution vulnerability that had lain dormant for 12 years since GitLab 7.0 in 2014. The article states that Calif Research researcher Ryan Austin, under OpenAI's Patch the Planet program, used large language models to locate PHP memory flaws, and that Wiz's Atlas multi-agent system also independently discovered the same set of vulnerabilities; it also mentions defensive-side efforts such as Microsoft Guardian Agents and Anthropic Project Glasswing, along with attack cases in which hackers used the ARTEX framework.

#Safety / alignment#Open-source ecosystem#English text is being prepared.

RRohan Paul21:38Opinion

AI score58

After OpenAI's Navier-Stokes announcement, Harvard mathematicians are divided over whether AI proofs require human understanding

After OpenAI's Navier-Stokes announcement, Harvard mathematicians are divided over whether a correct AI proof requires human understanding. Rohan Paul points out that a proof both shows a proposition to be true and teaches mathematicians reusable methods, whereas AI proofs often achieve only the former. Mathematicians value the certainty and technique of a proof; an AI proof that no one understands provides certainty but not technique.

#English text is being prepared.#Reasoning#OpenAI

CcnBeta21:09Opinion

AI score78

AI uncovers a PHP vulnerability hidden for 14 years, but the fix gets stuck in process

PHP was exposed for two remote code execution vulnerabilities rated 9.2, CVE-2026-109618 and CVE-2026-109619. OpenAI confirmed that the security firm Calif used GPT-5.6 Sol to discover the key vulnerability among them, and Wiz Research also independently reported the related issues. Details of the vulnerabilities leaked before patches were released, and PHP 8.1 and earlier versions cannot receive official updates, relying only on source patches or distribution backports. The PHP Foundation said that the first clue for CVE-2026-109618 appeared in an external security audit in 2024, and the prolonged failure to fix it was a structural failure.

#Safety / alignment#OpenAI#Anthropic

TThe Verge AI21:01Opinion

AI score56

AI shopping agents are ready, but some retailers are still on the defensive

A The Verge reporter tested shopping agents such as Meta Muse and found that Amazon has asked Meta to remove Amazon from Muse, and has previously sued Perplexity; Target's bot verification causes agents to connect only intermittently. The reporter also had Muse repeatedly bulk-add new items to cart on The RealReal to occupy the 20-minute reservation window, noting that this kind of automation squeezes out other buyers. Meta responded that Muse always asks for permission before purchasing and that retailer feedback has been generally positive.

#English text is being prepared.#Agents#OpenAI

AAlibaba Cloud Developer Community AI20:37Opinion

AI score64

How Playwright 1.60 Went from a Testing Tool to the Browser Operation Layer for AI Agents

The article argues that Playwright 1.60 has shifted from a testing framework to a runtime for Agents to operate browsers, that Test Agents (planner, generator, healer) have become production-ready, that the MCP server has matured further, and that ARIA snapshots now include bounding boxes. The author breaks this change down into four layers: perception, execution, diagnosis, and roles. ARIA snapshots provide structured page information; browser.bind() lets exploration, the test runner, and MCP share the same browser instance; traces are folded into HAR network capture; and planner, generator, and healer form a closed loop from plan to heal. The article also stresses that the healer does not touch assertions, and that fixes should be submitted as PR diffs for human review rather than merged automatically.

#English text is being prepared.#Agents#English text is being prepared.

336Kr · AI20:37Opinion

AI score60

Short Drama in 2026: AI Replaces Live-Action Crews and the Overseas Short Drama Ecosystem Is Reshuffled

Xiaguang Society reviewed the changes in the overseas short drama industry in 2026: multiple practitioners said that since May and June, live-action short drama projects have sharply decreased, with overseas live-action short drama shooting projects shrinking 30% to 50% year-on-year, while the cost of AI short dramas is about one-tenth that of live-action dramas. DataEye Research Institute data shows that Jiuzhou Culture's AI dramas and animated dramas already account for 95% of the company's total production capacity, with 1,399 new dramas produced and broadcast in March 2026 alone; Kunlun Wanwei's Tiangong AI's AI Native business had ARR exceeding US$800 million in Q2 2026, of which the AI short drama platform exceeded US$700 million. The article also records how the lowering of TikTok's short drama revenue-share coefficient caused a sharp drop in the amounts received by AI short drama overseas expansion companies, as well as the landscape in which Chinese teams, overseas transformation teams, and AI-native entrepreneurs are competing for new ecological niches.

#English text is being prepared.#Video#Industry news

336Kr · AI20:37Opinion

AI score66

MiniMax's Hong Kong shares fall below IPO price, with high revenue growth and share supply pressure coexisting

On October 22, MiniMax's Hong Kong shares closed at HK$160.7, falling below the IPO issue price, a maximum decline of 88% from the historical high of HK$1,330 in March this year. The article attributes the break below the issue price to the increase in share supply brought by the unlocking of about 153 million shares in July, the placement of 35.6 million new shares, and HK$6.5 billion in convertible bonds, while noting that the company's revenue in the first half of 2026 was US$116.6 million, up 283.1% year on year, R&D expenses were US$296.9 million, and net loss was US$358 million. The M3.1 Flash released this month ranks first among domestic models on the ALE benchmark, with online inference speed increased by nearly 3 times and the comprehensive usage cost of Agentic tasks reduced by 70%.

#Industry news#Model releases#English text is being prepared.

TTMTPost · AGI20:36Opinion

AI score60

ByteDance Builds an AI Video Tool Matrix: Jimeng, Xiaoyunque, Comic Drama Creation Tool, and Jianying Hub

ByteDance has formed an AI video tool matrix within its ecosystem covering Jimeng, Xiaoyunque, the Comic Drama Creation Tool, Dramagic, and Jianying Hub, segmented by scenarios such as premium film and television, short drama mass production, comic drama, and mass-market UGC. The Comic Drama Creation Tool, in beta since August, supports scripts of up to 200,000 characters and collaboration by up to 30 people, and can run 100 episodes at a time; on September 21, BytePlus launched the overseas short drama Agent Dramagic, while Jianying launched Jianying Hub, which connects AI generation and editing, at the Douyin Creator Conference. The article argues that this divergence stems from differences among content ecosystems such as Douyin, Hongguo, and PineDrama, and also aligns with the trend of AI tools evolving from general-purpose to vertical Agents.

#Industry news#Video#Agents

TTMTPost · AGI20:36Opinion

AI score63

Per-10,000-view revenue falls as low as 5 yuan: why the AI short drama business loses more the more it's made

The AI short drama industry has fallen into a predicament of losing more the more it's made. Industry per-10,000-view revenue has dropped from 30 to 100 yuan in the second half of last year to 5 to 10 yuan, a decline of nearly 90%. DataEye statistics show that in the first half of this year, 98.7% of AI micro short drama projects could not recoup their costs. The pure compute cost of one episode of a standard AI comic drama has risen from 17 yuan in the early days to 80 to 100 yuan, and traffic acquisition costs consume about 70% of total end-to-end costs. A small number of surviving teams have turned to B2B brand custom dramas or use AI comic dramas as customer acquisition bait, bypassing C-end revenue sharing.

#English text is being prepared.#Video#Industry news

TTMTPost · AGI20:36Opinion

AI score76

Anthropic prospectus exposed: revenue of $16.3 billion in half a year, book loss of $80.7 billion

The draft Anthropic IPO prospectus exposed by the media shows revenue of $16.3 billion in the first half of 2026, up nearly 14-fold year-over-year, with a net loss of $80.7 billion, of which about $72 billion comes from the revaluation of financing instruments such as convertible notes. Compute and infrastructure spending in the same period was nearly $9 billion, about 55% of revenue, and the company has also signed over $600 billion in forward compute commitments, which could reach up to $750 billion if additional agreements are finalized, with about 80% of these being irrevocable or take-or-pay arrangements. The article also mentions that of the $16.3 billion in revenue in the first half, more than $3.2 billion came from two major customers, and many major customers have not signed long-term purchase commitments.

Same story · 5 reports

#Industry news#OpenAI#Anthropic

TThe Guardian Technology20:30Opinion

AI score60

How the tech industry became a target in the US midterm elections

In the US midterm elections, there have already been more than 150 campaign ads focused on the data center or AI industry, with candidates from both parties tying their opponents to unpopular data center projects. The report states that the tech industry and its leaders have invested more than $500 million to influence the 2026 elections and ballot proposals, including $366 million from Musk, while Leading the Future, founded by OpenAI President Greg Brockman and Marc Andreessen, has raised more than $75 million. At least 20 gubernatorial candidates have called for a pause on data center construction. OpenAI has recently kept its distance from Leading the Future, and Brockman withdrew from a planned $25 million donation.

#English text is being prepared.#Industry news#English text is being prepared.

BBusiness Insider Tech18:49Opinion

AI score64

Bernstein Research: Only About 35% of Announced US AI Data Center Capacity Is Likely to Be Built

After analyzing more than 5,000 data center projects in North America, Bernstein Research estimates that only about 35% (roughly 170 gigawatts) of the nearly 500 gigawatts of announced capacity is likely to be built, with the rest classified as "bragawatts." Established operators such as Amazon, Google, Microsoft, and Meta received a perfect score of 5 on development qualifications, while O'Leary Digital, associated with Kevin O'Leary, scored only 0.35, with only about 489 megawatts of its 15-gigawatt plan expected to materialize. Aterio data shows that about 30 to 40 gigawatts of new capacity is announced each month, but only about 7 gigawatts breaks ground monthly and just 2 to 3 gigawatts enters operation.

#Industry news#English text is being prepared.#English text is being prepared.

WWallstreetcn17:47Opinion

AI score60

Wallstreetcn Reviews Silicon Valley's Chinese AI Entrepreneur Community: From HRT Interns to Big Tech's Core

Wallstreetcn reviewed the formation path of Silicon Valley's Chinese AI entrepreneur community, from Alexandr Wang, Scott Wu, and Jeff Yan, who interned a decade ago at the quantitative giant HRT, to today's leaders of Meta Superintelligence Labs, Cognition, and Hyperliquid, respectively. The article also lists chip chiefs such as Jensen Huang and Lisa Su, as well as core researchers at frontier labs such as Zhao Shengjia and Mark Chen, and cites a September 2026 Carnegie China report stating that 57% of the world's top AI researchers completed their undergraduate education at Chinese universities. The article also notes that Meta released Muse Wonder and built its most powerful version exclusively into its in-house Muse Code, squeezing the space for independent orchestration products like Cognition.

#Industry news#English text is being prepared.#Agents

336Kr · AI17:37Opinion

AI score83

PHP2Shell vulnerabilities found by AI take down government websites in multiple countries

Two vulnerabilities discovered by AI in the PHP core are collectively called PHP2Shell, both with an official severity score of 9.2; attackers may be able to remotely take control of servers without logging in, and institutions in Finland, the Netherlands, France, Germany and other countries took websites offline preventively around the patch release. The vulnerabilities were found by Calif using models provided through OpenAI's Patch the Planet program, while Wiz independently reproduced them at the same time using Gemini 4 Argon and GPT-5.6 Sol; within three weeks the two companies hit the same set of vulnerabilities from two directions. The article notes that AI has cut the cost of finding and verifying vulnerabilities from months for a top-tier team to one person for one week, and old-code vulnerabilities that were previously set aside because attacking them was too costly are now being exposed again.

#Safety / alignment#Agents#OpenAI

336Kr · AI17:37Opinion

AI score76

Personal agents launch in droves, most apps unwilling to open their entry points

Anthropic released a standalone mobile app Claude Duet on October 21, which can monitor email and calendars, chase bills, and place orders and make payments after user approval; it is based on Opus 5.5 and is free of extra charge during the preview period. The article compares similar personal agents such as Meta Muse and OpenAI dots, noting that they all have their own computers in the cloud and take over email, calendars, and wallets. Meanwhile, Amazon blocked Muse, Doubao's screen automation was rejected by more than 150 apps, and Walmart Agent Gateway and Doubao SAEP attempt to use protocols such as MCP to let apps proactively decide whether to allow agents.

Same story · 44 reports

#Product updates#Agents#Anthropic

HHuxiu17:17Opinion

AI score78

Anthropic IPO Draft Prospectus Revealed: $16.3 Billion in Revenue, $80.7 Billion Net Loss in Half a Year

Anthropic's draft IPO prospectus shows $16.3 billion in revenue and a net loss of $80.7 billion in the first half of 2026, of which $72 billion comes from the revaluation of financing instruments such as convertible notes and is not actual cash spending. Computing and infrastructure spending in the first half of the year was nearly $9 billion, about 55% of revenue, down from more than 200% in the same period last year, and the company had already begun to be profitable on an adjusted basis in the third quarter. The article points out that the real risk lies on the cost side: signed forward computing and infrastructure commitments exceed $600 billion and could reach as high as $750 billion; Reuters said that of the $518 billion at the time, about 80% was irrevocable or required minimum payments, while of the $16.3 billion in revenue in the first half of 2026, more than $3.2 billion came from just two major customers, and many major customers had not signed long-term purchase commitments.

#Industry news#English text is being prepared.#English text is being prepared.

EEast Money16:42Opinion

AI score58

Brokerages Debate AI Short Dramas: After Production Costs Fall, Where Does the Money Flow?

On October 21, the AI short drama and comic drama concepts continued to heat up in A-shares and Hong Kong stocks, with Yue Media hitting the daily limit for a second consecutive session, Wintime Culture hitting the daily limit, and Hong Kong-listed Duiba rising more than 40% intraday. Brokerages including Sinolink, Huafu, Huaxi, Zheshang, and China Merchants differ over the order of beneficiaries: Sinolink believes tool providers such as model invocation and Agent workflows may benefit first, Huafu emphasizes that upstream IP and downstream platforms will gain stronger bargaining power, while Huaxi points out that in the first half of 2026, the rate of new AI dramas on Douyin exceeding 100 million views was about 0.48%, and revenue per 10,000 views fell from about 30–100 yuan to 5–10 yuan. Several institutions also warn of risks including share prices being overpriced ahead of earnings and oversupply driving down unit monetization.

#Industry news#Video#Agents