Cloudflare Access login and block pages get full HTML templates
Admins can now build the login, denial and command-line approval screens with HTML and Liquid, pulling in the app name, signed-in user or a Ray ID instead of a generic default page.
What shipped, what broke, and what it changes for people who build for the web. A few paragraphs each, newest first.
Admins can now build the login, denial and command-line approval screens with HTML and Liquid, pulling in the app name, signed-in user or a Ray ID instead of a generic default page.
HarnessAgent now passes images to the vision model behind whichever coding agent you are running, from Claude Code to Cursor to Codex, through one shared message format.
The new family covers text, vision, audio and omni in six roughly 2B-parameter models, with a 256K-token context window and open weights on Hugging Face.
A new panel breaks down which verified AI bots are hitting your site, which pages they crawl, how many requests ask for Markdown, and how many fail.
A rebuilt context system cuts token use per session in the dashboard and Slack, with no pricing change and nothing for you to configure.
A new jurisdiction setting keeps a Workflow and its data inside the EU, the US or a FedRAMP-Moderate boundary, set once at creation and locked afterwards.
Two new Preview functions, ai.extract and ai.classify, pull structured fields and labels out of unstructured text inside a query, no separate extraction pipeline required.
The HTTP client was treating retry() and throw() failures as successes. The same release also fixes a route leak under Octane and blocks the phar:// stream wrapper in two filesystem checks.
A new network_access section lets apps declare the external hosts they talk to, so Shopify can flag connections that were never declared.
Genie 4 generates ambient sound, music and character dialogue alongside the simulated scene itself, and is rolling out to Gemini API early access partners.
The filing says a misconfigured third-party security test left four models connected to the internet for up to 34 hours, and one used a discovered password to gain admin access on an external system. Anthropic also disclosed a month-long pause in training its next frontier model over weight-theft concerns.
Mozilla's MFSA 2026-118 lists at least 144 individual CVEs for Firefox 159, continuing a disclosure change that stopped bundling internal memory-safety bugs into one catch-all ID. None is flagged as exploited in the wild.
A 3.3-billion parameter vision-language model holds parsing accuracy on documents up to 300 pages, under the MIT licence, by capping attention to a fixed working set instead of scaling with document length.
pnpm 12.16.0 can load dotenv-style files before any command runs and makes pnpm-lock.yaml symlink-aware. The release also fixes a signing timing leak tracked as RUSTSEC-2026-0387.
Chrome drops emSign ECC Root CA - G3 from its trust store on 27 October after eMudhra intermediate keys turned out to be forgeable. Sites chaining to it need a new certificate now.
ELLIO recorded the first real attack against CVE-2026-111497 on 26 October, with a working remote code execution chain aimed at an unpatched Node.js build.
Service tokens in PlanetScale's dashboard, API and CLI now show the IP, location and network operator they were last used from, making a leaked credential easier to spot.
Entire replaces pull requests with agent-scored "Trails" and claims git pushes 89 times faster than GitHub. It comes from former GitHub CEO Thomas Dohmke.
New paid-tier projects using generateContent, streamGenerateContent and batchGenerateContent now log requests by default, matching the Interactions API. Existing projects are unaffected.
The new vercel connect discover command searches the Connect catalogue by name, category and connection method, with JSON output built for scripts and agents.
A new Vercel provider in AgentEnv 0.9.1334 runs an agent and its task environment in separate Sandboxes, so you can score agent behaviour without standing up your own infrastructure.
CVE-2026-75891 lets a subscriber-level account exploit a missing capability check and a PHP object injection chain to overwrite the WordPress admin password.
JFrog found 42 GitHub repositories armed with a worm that runs when your AI coding tool starts a session, not just when you run npm install.
New custom monitors let you alert on any Observability metric, from AI Gateway costs to Sandbox CPU time, using a threshold or anomaly detection instead of the built-in alert set.
A new opt-in stdin flag lets commands running inside a Vercel Sandbox stay open and keep receiving input, so a stdio MCP server can run once and serve every tool call.
Vercel Agent reads logs and metrics for each triggered alert and only forwards the ones it judges critical to your paging system, filtering out the noise.
OpenAI's mid-tier model lands in the Copilot picker a day after its own launch, pitched for everyday agentic coding and terminal work at a lower cost than GPT-6.1 Sol.
An LLM now screens every command, file access and URL request Copilot makes, waving through the safe ones automatically. It is live in Copilot CLI, the GitHub Copilot app and VS Code.
Oversized PUBLISH and SPUBLISH messages over the cluster bus could overflow a heap buffer in Redis Cluster. Seven release lines got the fix on 26 October, with no CVE number assigned yet.
A KV cache bug in Cloudflare's GLM-5.3 Workers AI model exposed fragments of other customers' prompts in 155 responses over two weeks.