RC RANDOM CHAOS

AI security

18 posts

Gemini 3.8 Live broke two security assumptions
Article

Gemini 3.8 Live broke two security assumptions

Gemini 3.8 Live runs tools in the background and ships production speech-to-speech, breaking two voice-agent security assumptions engineers rely on.

One attacker, three rivals, zero coincidence.
Article

One attacker, three rivals, zero coincidence.

One Israeli firm, Irregular, ran the red-team exercises behind the OpenAI, Anthropic, and Meta incidents where AI models broke into real systems.

DayBreak doesn't make your systems vulnerable
Article

DayBreak doesn't make your systems vulnerable

OpenAI's Daybreak bundles Codex Security, the GPT-5.5-Cyber model, and Patch the Planet to shift AI effort from finding vulnerabilities to patching them.

GLM-5.2 binds to a port and opens a shell
Article

GLM-5.2 binds to a port and opens a shell

Unsloth Studio runs GLM-5.2 on a local port with built-in Python and Bash execution, plus a Cloudflare tunnel option that can expose that shell online.

The door was unlocked, not picked
Article

The door was unlocked, not picked

A three-word prompt, "fix this code," triggered a US export ban on Anthropic's Fable 5 and Mythos 5, per Katie Moussouris's read of the research.

Meta's chatbot worked exactly as designed.
Article

Meta's chatbot worked exactly as designed.

Meta says at least 20,225 Instagram accounts were hijacked through an AI chatbot that reset passwords without verifying the account's email address.