Today's Headlines

  • PHP2Shell exploitation was attempted four days before disclosure, and the NSA advises taking unpatchable servers offline
  • DeepSeek is reported to be raising about 100 billion yuan, two days after a US exclusion order
  • Anthropic donates the Agent Skills standard to a Linux Foundation body
  • Meta creates a non-voting class of stock to help fund its AI spending

The day's thread is things that kept moving after someone tried to hold them back. The critical PHP flaws found with AI were meant to go public together with their fixes. Amazon, however, saw an exploitation attempt four days before disclosure, and says the flaws are now broadly exploited.

Money kept moving too. Two days after the US ordered federal civilian agencies to remove DeepSeek products, reports said DeepSeek is raising about 100 billion yuan.

Today's Top Three

PHP2Shell exploitation was attempted four days before disclosure, and the NSA advises taking unpatchable servers offline

Amazon Threat Intelligence said on October 21 that it had seen an attempt to exploit the critical PHP flaws known as PHP2Shell on October 17, four days before they were made public.

PHP2Shell is the name for two vulnerabilities, CVE-2026-109618 and CVE-2026-109619. They were found with the help of AI, and PHP's maintainers released fixes on October 21. The flaws can let an attacker take over a web server without logging in.

According to Amazon, the October 17 attempt came from an IP address tied to a commercial anonymization network and hit its MadPot honeypot network. On disclosure day, Amazon found several proof-of-concept exploits even before the embargo lifted, and new exploits were used against its sensors within hours after it did. Amazon says PHP2Shell is now being broadly exploited in the wild.

Amazon advises starting investigations from October 17 rather than from the disclosure date. It argues that AI-assisted analysis has made it cheaper to turn a patch into a working exploit, so patch timelines for critical internet-facing systems should be measured in hours.

The US National Security Agency (NSA) also urged organizations using PHP on October 21 to prioritize the patches. Where an immediate update is not feasible, it recommends taking the system offline until the flaws are fully mitigated.

DeepSeek is reported to be raising about 100 billion yuan, two days after a US exclusion order

Chinese AI company DeepSeek is raising about 100 billion yuan (about $15 billion), Bloomberg and Reuters reported.

Bloomberg reported that DeepSeek had raised the money, while Reuters, citing two people familiar with the matter, said the round was expected to close in the coming days. Tencent and battery maker CATL are among the investors, according to the reports.

DeepSeek raised its target at least twice, from 50 billion yuan to 80 billion yuan and then to about 100 billion yuan, according to Bloomberg's report as relayed by China Daily. The round would rank among China's largest.

The reports came two days after the US Department of Homeland Security ordered federal civilian agencies to remove DeepSeek products. According to TechCrunch, DeepSeek also raised more than $7 billion at a $50 billion valuation in June.

Anthropic donates the Agent Skills standard to a Linux Foundation body, following MCP

Anthropic announced on October 22 that it is donating Agent Skills, the open standard that gives AI agents specialized knowledge and workflows, to the Agentic AI Foundation (AAIF) under the Linux Foundation.

A skill is a folder with a SKILL.md file that teaches an agent a specific task or domain, plus any scripts or reference material it needs. Anthropic launched skills with Claude Code in October 2025 and published Agent Skills as an open standard that December.

More than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot and VS Code, have integrated Agent Skills, according to Anthropic. One million skills have been published so far, with nearly 280 million installs.

Anthropic is donating the name and logo, the GitHub organization and the agentskills.io domain. It co-founded the AAIF with Block and OpenAI in 2025 and earlier donated the Model Context Protocol (MCP) to it. Anyone can use or contribute to the standard without joining the AAIF, and engineers from Anthropic, Google, OpenAI and other companies will maintain it as individuals.

Anthropic has proposed a roadmap that starts with versioned releases, a formal proposal process and security guidance for people who write and install skills.

Other Developments

Models

  • OpenAI made its Decisions API generally available on October 22. Built on GPT-6 Luna, it answers questions with bounded answers up to 10 times faster than GPT-6 Luna through the Responses API, and a single request can now carry up to 400 questions, twice the previous limit. Build more responsive apps with the Decisions API (OpenAI)

Security

Research

Products

Policy

Business

  • Meta announced with its third-quarter results on October 21 that it will create a new class of publicly listed, non-voting Class C common stock for capital raising, employee compensation and acquisitions, effective January 19, 2027. It cited the scale of investment required amid the AI transformation, and will distribute one Class C share for each existing share as a stock dividend. Quarterly revenue rose 26% year over year to $64.6 billion. Meta Reports Third Quarter 2026 Results (Meta)
  • Uber began rolling out OpenAI's ChatGPT Work company-wide on October 22, making the agent available to more than 30,000 employees. The rollout expands a pilot of 1,000 employees, about 80% of whom were non-engineers, according to TechCrunch. Uber brings ChatGPT Work to all employees with OpenAI (OpenAI)

Source: Selected by the editors from the AI news inbox (collected October 23, 2026: 70 items, 15 primary and 55 secondary).