Expert perspectives on AI security, fraud detection, KYC compliance, and the evolving AI landscape.
A practical comparison of SCA and SAST platforms for teams building and shipping AI-assisted code, across PR workflow, dependency risk, code scanning, secrets, and licensing.
Read Full Article โChoosing "don't ask again" in Claude Code can save a password into a settings file. 1,122 public repositories have committed one. Most secret scanners miss passwords typed into commands, and so did ours.
Read Article โA Claude Code security agent and Vybscan on five codebases with answer keys: what each finds, what each costs over a month, and what an agent forgets between runs.
Read Article โThe ponytail-sec Claude Code plugin and Vybscan on five codebases with answer keys: recall, precision and cost, both triaged the same way, and the three bugs it found that we missed.
Read Article โVybscan is built so that your code stays yours: we read it to scan it, and we don't store it. How we protect your code, and where we stand on SOC 2 and CSA STAR.
Read Article โNine things attackers want, the code weaknesses that give it to them, and real CVEs in Gogs, DataEase, Kanboard and Umbraco that show each one.
Read Article โPHP2Shell, GitLab, Node.js, LiteLLM and more, most of it months old. Without a code scanner you learn from the news and search by hand. What to do this week, and what a scanner changes.
Read Article โVybscan on the industry's standard test apps, from the OWASP Benchmark to Juice Shop, PyGoat, RailsGoat, WebGoat.NET, DVWA and TerraGoat, scored with each project's own answer key.
Read Article โOn ZeroPath's own benchmark, Vybscan found 82.9% of technical vulnerabilities to ZeroPath's 80.0%, and matched it on business logic.
Read Article โPrecision, recall and cost for ten scanners on the RealVuln benchmark, corrected for its missing labels. Plus 100% SCA recall on three public codebases.
Read Article โA scanner that misses a bug costs more than one that raises a false alarm. How we build and measure for recall, and our numbers on three codebases.
Read Article โA frontier model finds security bugs in small repos, and sometimes beats us. Four things it cannot do, and our own evidence against us.
Read Article โAI models keep inventing the same package names. Attackers register them. Two documented cases, and what actually stops it.
Read Article โImpersonation, AI answers read aloud, off-mic coaching, cloned voices. Proctoring watches the room and misses most of them.
Read Article โA Claude model ran a dependency-confusion attack during a security test and 15 machines executed the payload. How scanning dependencies on every pull request stops it.
Read Article โThe real per-call cost of Gemini 2.5 Pro, Qwen 72B on OpenRouter, and Qwen 7B local for fraud detection, once the credits run out. Why the cheapest model is not always the right one.
Read Article โUpdated for 2026: Qwen2.5-VL, InternVL2.5, Phi-4 Vision, Moondream, and the latest cloud APIs compared for production tamper detection workflows.
Read Article โEvery AI term decoded through two lenses: the human situation it mirrors and the pre-AI software concept it replaced. Your field guide to the new vocabulary.
Read Article โCore security risks of OpenClawForCISO agents and hardening best practices for safe deployment.
Read Article โOur Bayesian KYCC engine doesn't just detect fake images. It explains why an image is risky, combining visual AI with behaviour and provenance signals.
Read Article โA classification of software applications in the AI era, from AI-core business logic to AI-assisted and AI-generated systems.
Read Article โFrom the Bocca della Veritร to Solomon's wisdom, the ancient quest for truth verification and why AI has made it 10x harder.
Read Article โAI has inherited all existing security issues and introduced new ones: adversarial attacks, data poisoning, deepfakes, and systemic risks.
Read Article โGet our weekly digest of cybersecurity news, compliance updates, and legal tech insights delivered to your inbox.
No spam. Unsubscribe anytime.