GitHub launched public-preview code scanning for AI apps and agents on 28 Oct 2026, flagging indirect prompt injection, data disclosure and excessive agency in PRs.
On 28 Oct 2026 Cisco open-sourced Project Quarry, a detection language and engine for multi-step attacks on AI agents, tool use and MCP.
Binance said on 29 Oct 2026 a security incident impacted about $6.3B in assets; ~$3B has been frozen, and TRM Labs and Chainalysis attribute the attack to North Korea's TraderTraitor.
Malicious LiteLLM 1.107.0 hit PyPI and Docker, stealing secrets. Roll back to 1.106.2 and rotate credentials.
METR and Redwood Research’s pilot access found novel flaws in OpenAI’s misalignment monitors, including hiding hundreds of agent actions.
PHP shipped 8.5.12–8.2.35 to fix CVE-2026-109618 and CVE-2026-109619 (PHP2Shell). WordPress says update PHP, not WordPress core.
Replit CEO Amjad Masad said at TechCrunch Disrupt that OpenAI is now the coding platform's preferred partner and default for Replit Agent, though Claude models remain available.
A fake app called WhatsApp Messenger has been downloaded more than 500,000 times on Google Play and reportedly appeared in sponsored results above the real app.
Next.js shipped an out-of-band security update on October 14 fixing two critical remote code execution flaws and a high-severity DoS bug. Upgrade to 16.4.1 or 15.5.28 now.
Chatbot startup Character.AI has filed for Chapter 11 bankruptcy, with Disney offering $20 million for a license to part of its technology.
Apple updated its App Review Guidelines to target nonconsensual intimate imagery, including AI-generated content, and apps that change their behavior after approval.
Meta has released the weights of Muse Spark Open, an 816-billion-parameter model that, with help from Apple's MLX framework, can run locally on two M5 Ultra Mac Studios.
Apps built with Google AI Studio can now be installed on a user's own Android device through Google Play without being published publicly.
Microsoft's third-generation Maia chip will become the main inference system for its products from 2027, promising 40% more tokens per megawatt than leading GPU systems.
Google has released Gemini 4 Argon worldwide for developers, enterprises and paying subscribers, targeting complex coding, science and agentic work.