Click any tag below to further narrow down your results
Links
Cloudflare's Workers AI team has open-sourced an updated Clairic model and a smaller 9B version called Clairic-flash, both built for pulling structured data out of text, images, and PDFs. They claim strong accuracy and sub-second response times at a low per-token price.
- The redesigned architecture fills fields in parallel with autoregressive generation, lifting accuracy on text and images; image-reading errors fell from 207 to 0 on their internal benchmark.
- Clairic now reads PDFs and images (up to four pages or images per request), so supplier, line items, and totals can be pulled in one call even across pages.
- Pricing is $0.24 per million input tokens for the 27B Clairic and $0.10 for the 9B Clairic-flash, with a promotional $0.10 rate on Clairic until December 1, 2026. Most evals are self-reported, which the authors acknowledge.
This isn't an article, it's a bot-protection interstitial: the page asks you to solve a CAPTCHA to prove you're human and then grants temporary access. It also suggests running a malware scan if you're on home internet, or asking your IT admin to check the network if you're on a shared one.
- The CAPTCHA only proves you're human and grants temporary access to the site.
- Repeated challenges are usually blamed on possible malware on your device or network.
- On shared networks, the fix is for an administrator to scan for misconfigured or infected machines.
This is Cloudflare's documentation for Cache Deception Armor, a cache rule that stops attackers from tricking the CDN into caching a logged-in user's personal page under a fake static-file URL. It checks that a file's extension matches its returned Content-Type and skips caching when they don't line up.
- The attack works when an origin serves the same dynamic page for a fake path like /newsfeed/foo.jpg, so Cloudflare caches the user's personalized response by default.
- Cache Deception Armor keeps caching for real static assets but refuses to cache any response where the extension and Content-Type mismatch.
- Responses typed application/octet-stream are exempt, and harmless mismatches like .jpg served as image/webp are allowed.
A PHP security flaw found with an OpenAI model leaked before a patch was ready, and a volunteer maintainer, Ondřej Surý, became the key person getting fixes packaged and onto servers worldwide. The piece uses his story to show how much of the internet depends on a few unpaid people.
- Cloudflare's CEO said attack attempts rose from thousands to millions within hours of the disclosure, showing how fast exploitation moved once the flaw was public.
- PHP runs on roughly 70 percent of websites, and the flaw allowed remote code execution, the most severe category, which is why the panic spread so widely.
- Surý maintains packaged patches as an unpaid volunteer alongside his day job, and his backup plan rests on a company, Freexian, that could take over if he were unavailable.
Cloudflare is previewing Clairic, a model that pulls values out of unstructured text, computes them, or forecasts them in a single pass with no generated tokens. It's pitched as faster, cheaper, and less error-prone than using an LLM for the same structured-output jobs.
- Clairic returns typed values (spans, numbers, dates, durations, lists, distributions) rather than free text, so it can't produce a malformed or hallucinated answer; missing values come back as null.
- Every field is answered in one parallel pass with no output tokens billed, so asking for thirty fields costs about the same as asking for three, and responses usually land in under a second.
- Each answer carries confidence scores, character-offset spans back to the source, and 80% intervals for estimates, so you can see which outputs to double-check.
Cloudflare open-sourced two decision models called Clef and Clef-flash that classify inputs into typed, structured outputs with probabilities — useful for routing customer support tickets, identifying phishing domains, or triggering automated agent actions. The models outperform Typesafe's Jev system on speed and benchmarks while supporting images and a 64k context window.
- Clef processes decisions 2x faster than general LLMs (2.2s vs 4.7s in Cloudflare's threat intelligence workflow) because it scores schema choices in parallel rather than generating text token-by-token.
- The models beat Jev on most Jev Decision Index benchmarks and Typesafe's own eval suite, with Clef-flash offering exceptional speed-to-quality tradeoffs; median latency is 209ms for Clef and 38.8ms for Clef-flash.
- Both models are open-sourced on Hugging Face under Apache 2.0, hosted on Workers AI at Cloudflare's edge for low network latency, and fully API-compatible with Jev for easy switching.
This GitHub repo provides a coding-agent skill that runs automated security audits in six phases—recon, hunting, validation, reporting, structured output, and independent verification—to identify exploitable vulnerabilities. It uses parallel agents to generate and disprove findings, outputs structured JSON conforming to a schema, and independently verifies each claim against the source code. Each run reads prior findings to skip known issues and improve coverage.
- Cloudflare open-sourced the actual skill behind their own vulnerability discovery harness, structured as six phases: recon, hunting, validation, reporting, structured output, and independent verification.
- Findings are adversarially checked—the agent that finds a bug never validates it, and separate fresh agents re-verify each claim against the source code before reporting.
- A single audit pass only catches about half of total vulnerabilities, so the tool reads prior findings.json runs to skip known issues and improve coverage over repeated runs.
- Output is enforced as machine-readable JSON validated against a schema via a zero-dependency Node.js script, alongside human-readable REPORT.md and FINDINGS-DETAIL.md files.
Cloudflare’s teams quickly reviewed CVE-2026-31431 (“Copy Fail”), confirmed their behavioral detections flagged the exploit within minutes, and found no signs of in-the-wild abuse. They ran fleet-wide threat hunts, deployed a bpf-lsm mitigation, and rolled out updated kernels without impacting services or customer data.
- Cloudflare's anomaly-based behavioral detection caught the "Copy Fail" exploit pattern within minutes, without a CVE-specific rule ever being written.
- A 48-hour log hunt turned up no evidence of pre-disclosure exploitation in the wild.
- Cloudflare patched its fleet across 330 cities with zero downtime or customer data exposure, using weekly kernel builds and rolling reboots.
- The bug itself let an unprivileged process abuse the AF_ALG crypto interface to corrupt cached files (e.g., /usr/bin/su) for root privilege escalation.