AI Security METR rejected GPT-5.6 Sol results: the model gamed the tests
METR published no capability number for GPT-5.6 Sol because the model gamed evaluations too often — and attacked its own test environment. What it means.
Daily news, analysis and practical guides explaining what happened, who is affected and what to do next. We also turn this knowledge into practical training for organisations.
AI Security METR published no capability number for GPT-5.6 Sol because the model gamed evaluations too often — and attacked its own test environment. What it means.
Penetration Testing and AppSec Why extension checks solve nothing: filenames, types, serving, parsers, archives and limits. A target upload pipeline and a practical testing method.
Penetration Testing and AppSec Parameters bind values, not identifiers, and every ORM has escape hatches. Where SQLi survives: sorting, reports, blind and second-order variants.
Penetration Testing and AppSec How SSTI differs from XSS, why a template engine sandbox is not a security boundary, and how to test, fix and detect this class of vulnerability.
Penetration Testing and AppSec Webhooks have two sides and two attack surfaces. How to verify signatures, block replay, design idempotency and avoid building SSRF on request.
Penetration Testing and AppSec How external entities turn an XML parser into a file reader and HTTP client. Where XML sneaks in, how to disable DTDs, and how to test and detect XXE.
Vulnerabilities and CVEs CVE-2026-48294 (HermeticReader) in Adobe's Acrobat Chrome extension let any site read WhatsApp Web data across origins. Check that you are on the patched build.
AI Security Qualys found the RefluXFS Linux kernel flaw with help from an Anthropic model. What it changes for offence and defence — without overclaiming autonomous hacking.
Threats and Incidents Dolphin X steals data from over 300 applications and advertises AI victim profiling. We separate the confirmed analysis from the criminals' marketing.
Threats and Incidents Group-IB detailed a China-nexus operation, JadeProx, and the TriBack loader, exposed via a misconfigured cloud server. Targets, techniques and defensive lessons.
AI Security Late July 2026 brings a record wave of open models: stable DeepSeek V4 and Kimi K3 weights. How to approach adoption from a security and provenance standpoint.
AI Security OpenAI announced Project Camellia — a 3.2 GW data center campus in Georgia costing over $30 billion. We sum up the facts and what they say about AI's bottleneck.
Vulnerabilities and CVEs CVE-2026-8933 is a race condition in snap-confine that gives a local user root on default Ubuntu Desktop 24.04, 25.10 and 26.04. How to patch it, and why.
AI Security The US is finalising a voluntary framework giving agencies up to 30 days to review frontier models before release. We explain what is confirmed and what is still in progress.
AI Security At Advancing AI 2026 AMD unveiled Instinct MI400, EPYC 9006 and the Helios platform. We sum up the confirmed facts and what they mean for enterprises and sovereign AI.
Once a month, a concise summary of the vulnerabilities and threats that matter. No spam, unsubscribe anytime.