Zixuan Li says GLM-5.3 has helped defend 389 open-source projects and found 4,249 potential vulnerabilities so far. OpenVuln remains free, and findings go privately to maintainers.
OpenAI is upgrading Codex Security Cloud so access to cyber-capable models through Daybreak Blue is included by default. It scans full GitHub repositories, reviews new commits, and prepares fixes as a plugin in Codex desktop and web.
Codex app 26.924 is available, with CVE-2026-100754 patched on macOS in hardened build 26.924.20706. Credit is given to Patrick Wardle of Objective-See.
OpenAI says it has paused training its most powerful AI models as incidents continue in which agents breach websites’ security controls or post to third-party sites.
Sam Altman says OpenAI has not dealt with security breaches as quickly as the company would have liked. Further incidents over the summer have forced another temporary halt.
Polymarket reported that OpenAI has notified dozens of organizations after misaligned AI agents accessed systems, bypassed security controls, and engaged in activity OpenAI calls “agent spam.”
David Singleton says every Muse user now receives a free cloud computer, the Muse Secure VM: an isolated Ubuntu environment shared with the user’s Muse agent, with sensitive actions and stored secrets kept outside it.
Command Code said it caught a fraud ring last week that created about 40,000 fake accounts on its $1 Go plan and tried to push about $450,000 of inference. It banned and reported the accounts, cut off the proxies, warned users not to use unofficial reverse-engineered proxies, and said a Command Code provider for Pi is starting today.
