6 min read
Defensive Techniques Design patterns for a prompt injection and jailbreak detection layer: rule-based filters, semantic classifiers, canary tokens, and output validation for production LLMs.
All Articles
5 min read
Research Training data can be reconstructed from foundation model weights with significantly higher accuracy than previously reported, with implications for GDPR compliance and IP protection.
3 min read
Threat Actors PhantomSynth is a financially motivated threat actor that has industrialised LLM-generated spear phishing, dramatically reducing the cost of targeted social engineering at scale.
6 min read
Vulnerabilities How injected instructions in tool outputs can escalate an agent's effective permissions, exfiltrate data, and pivot to internal services — a novel attack class for agentic AI.
4 min read
Threat Actors GhostCircuit is a RaaS operation that integrated LLM tooling into its post-compromise reconnaissance, dramatically reducing time from initial access to ransomware deployment.