Researchers from Zhiyi Mou et al. have published an attack framework showing that persona conditioning can weaponize LLMs' compulsion to maintain character, generating up to 207x token amplification — turning a standard API call into an economic availability attack against any service with per-token billing.
Academic research has documented multiple reliable techniques for stripping or spoofing LLM output watermarks. With EU AI Act Article 50 enforcement arriving in August 2026, the gap between compliance theater and actual detection capability is about to matter.
Check Point Research's annual AI security report documents a fundamental shift: AI is now executing attack chains autonomously, not just planning them. The numbers are stark.
Researchers from Toronto, Cambridge, and ServiceNow demonstrated an AI worm that ingests public vulnerability advisories at runtime and synthesises working exploits for CVEs it was never trained on, successfully compromising targets across a simulated network.
Threat actors embed prompt injection payloads in third-party LLM plugins and data sources to hijack AI agent actions, exfiltrate data, and pivot within enterprise environments.
LLMs with large context windows can be reliably jailbroken by embedding hundreds of fictitious dialogues before the target request. The attack scales with context length.
Post-mortem of a multi-stage intrusion using LLM-generated spear phishing, AI-assisted credential stuffing, and automated recon to compromise a mid-market wealth management firm.
Model inversion and training data extraction attacks allow adversaries to recover PII, proprietary data, and trade secrets from fine-tuned LLMs exposed via API.