Chinese AI models: evaluate cost and security before switching
Lower AI prices create more options. An enterprise decision needs evidence about outcomes, data exposure and the ability to change providers safely.
Blog
Praktyczne poradniki o pentestach, badaniu podatności i narzędziach bezpieczeństwa opartych na AI.
Artykuły są publikowane po angielsku i portugalsku.
AI i narzędzia
22 wpisy
AI replication on IoT: what to test after the Anthropic study
The study investigates capabilities in a simplified environment. Pentesty’s analysis turns the discussion into verifiable questions about connected devices, agent autonomy and access boundaries.
AI governance: connect costs, permissions and evidence of use
A spending dashboard cannot explain AI use on its own. Connect identity, purpose, budget and outcomes so deviations can be investigated.
DeepSeek context compression: the reliability test missing from your pilot
Average accuracy can hide retrieval failures. Turn the DeepSeek study into a useful evaluation for enterprise systems and their evidence requirements.
Meta VR Glasses: enterprise security and privacy
What to assess before bringing virtual reality to work: data capture, accounts, AI and APIs. Check the announcement details and Pentesty's original checklist for an enterprise pilot.
OpenAI Safety Guardrails: What to Test Before Trusting an AI Agent
Six misalignment reports raise a practical security question: what can an agent actually do? Explore a reusable checklist for permissions, data exposure, memory and independent evidence.
The Future of Security Compliance Is AI-Native
Y Combinator named AI-Native Compliance Infrastructure a Fall 2026 Request for Startups. Why pentesting is the workflow most ready for that shift, how continuous assessment turns SOC 2, ISO 27001 and PCI DSS evidence into a byproduct instead of a project, and what changes when the security loop runs at the speed of deploys.
Frontier AI Just Got Better at Hacking. Here's What That Means for Your Security Posture
OpenAI previewed GPT-5.6 Sol, their most capable model for cybersecurity yet. Exploitation primitives in Chromium and Firefox, U.S. government gating, and why the attacker-defender symmetry means continuous testing is no longer optional.
SpaceX Acquires Cursor for $60 Billion: What It Means for Software Security
SpaceX's all-stock acquisition of Anysphere reshapes the software supply chain at a global scale. Known CVEs, prompt injection vectors, developer telemetry risks, and a hardening checklist for every team running Cursor in production.