
OpenAI: Training Runs Were Never Monitored
OpenAI's chief research officer admits monitors were never on during training. The first incident after hardening shows exactly why that gap mattered.
> booting the archive…
we use a small set of cookies. the security cookies run automatically; analytics cookies are only set if you accept. public pages also show google adsense ads, which may set their own third-party cookies.
> booting the archive…
Threats, privacy, zero-trust, and digital rights. — 82 transmissions

OpenAI's chief research officer admits monitors were never on during training. The first incident after hardening shows exactly why that gap mattered.

Cloudflare plans a public CA with GlobalSign root material and Merkle Tree Certificates from Q1 2027 — post-quantum TLS that still works on old devices.

Microsoft ties NeedyMantis to China-linked Storm-3069 — DLL sideloading, custom archives and WebSockets C2 built to outlast the initial breach.

Park24 confirmed 6.6 million Times Car accounts were taken, licence images included. This is a data-model leak, not a credential leak.

NVIDIA split agent safety into two layers: OpenShell sandboxes the agent, Sentry watches it from BlueField-4 silicon. The split is the actual idea.

Two NetScaler zero-days are now in CISA's exploited catalog with a 30 September deadline, and installing the fix can erase the evidence you will need.

Zenity Labs found three flaws in Salesforce Agentforce where a poisoned Web-to-Lead form leaked CRM data over DNS, with no login and not one click.

A hidden Elementor experiment let a query string switch off WordPress REST CSRF protection, so one clicked link could mint a new admin. Patched in 4.3.2.

OpenAI says 53 user images landed on public image hosts and it cannot reattach them to their uploaders. Consent to train is not consent to republish.

third-party.com, the docs placeholder everyone copied, now serves ClickFix; 13 lookalike domains carry scams. Static checks miss all of it — audit your docs.

CARBONATO hijacks exposed Docker hosts, installs Hermes Agent, rewrites its persona, then drives them via Telegram. AI does post-exploitation, scripts spread.

GitLab's "Email work item" address hides an account-wide token. Leak it and anyone can push to main, run CI jobs as you, and skip IP allowlists.