AI firms expand Singapore office footprints
OpenAI was said to be negotiating for 100,000 square feet at Shaw Tower as Anthropic, Mistral AI, Sierra and Manus also added office space in Singapore.
Read briefArtificial Intelligence
07OpenAI Codex sandbox escapes let commands run on host
Researchers found two ways out of OpenAI Codex’s sandbox, including one that could trigger host commands from read-only mode, and OpenAI patched both within eight days of being told.
Read briefAnthropic delays IPO to November
Anthropic reportedly shifted its planned IPO from October to November 2026 as advisers wanted stronger third-quarter results, while investors expected a $2 trillion valuation and the company faced funding and security questions.
Read briefHacktron Uses Claude to Take Over OpenAI Accounts
Three Hacktron researchers linked a Discourse image flaw and an OpenAI login weakness to reach staff accounts, then reported the bugs, showed limited access and stopped.
Read briefTrump plans AI czar and AI Force
Donald Trump said he would appoint an artificial intelligence czar and create an AI Force, while dismissing safety fears as a hoax and promising not to hold back the industry.
Read briefQwen3.8-Omni-Flash nears Gemini Flash benchmarks at lower cost
Qwen said its new multimodal model for AI agents handled audio and video together, used tools independently, and was priced below Gemini 3.8 Flash on cited benchmarks.
Read briefUnity releases plugins for Claude Code and Codex
Unity released official plugins for Claude Code and OpenAI’s Codex, with skills maintained by Unity teams and aimed at reducing reliance on older tutorials and forum posts for engine work.
Read briefCybersecurity
04Malicious npm packages bypass install-script checks at runtime
Checkmarx said an npm campaign centred on indexed-btree hid its loader in normal runtime code, let the package appear clean at install, and used extra packages to spread the operation.
Read briefShinyHunters breaches Clop leak site, threatens extortion
ShinyHunters said it used a Grav CMS upload flaw to deface Clop's Tor leak site and steal data, while BleepingComputer confirmed the defacement but not the broader theft claims.
Read briefOrchid Security publishes identity visibility guide
The article defined identity visibility, contrasted it with IAM policy and runtime behaviour, and sketched the capabilities and vendors it said organisations should consider for cloud and multicloud environments.
Read briefSolarWinds patches Access Rights Manager flaw
SolarWinds issued updates for Access Rights Manager after a reported hard-coded key flaw that could allow unauthenticated remote code execution, and it said the issue was fixed in ARM 2026.2.1.
Read brief