Eresus research, advisory, and security news
We collect writing, advisories, and current-event analysis around AI security, the MCP ecosystem, application security, and real attack chaining here.
Latest Posts
Strengthening AI Security with Eresus Security Defense Services
As organizations rapidly adopt generative AI, they face a new frontier of security challenges that traditional testing approaches simply cannot address.
Llama 4 Series Vulnerability Assessment: Scout vs. Maverick
Meta has launched the Llama 4 family, featuring models built on a mixture-of-experts (MoE) architecture. Here is our vulnerability assessment.
AI Risk Report: Fast-Growing Threats in AI Runtime
A comprehensive look into the fast-growing vulnerabilities affecting AI systems in runtime environments, featuring Eresus Sentinel.
The Cost of Being Wordy: Detecting Resource-Draining Prompts
The breakthrough of LLMs has captivated the NLP world, but resource-draining prompts can cause Denial of Wallet (DoW) attacks in cloud models.
Tools and Technologies for Secure by Design AI Systems
Examining the specialized tools and technologies needed to secure non-deterministic, probabilistic AI systems effectively.
Securing Agentic AI: Where MLSecOps Meets DevSecOps
Understanding Agentic AI systems that go beyond traditional AI models by acting autonomously with limited human oversight.
The Evolution of AI Security: Why Secure by Design Matters
Protecting AI systems requires a fundamental shift in security thinking. An intro to the Secure By Design framework applied to AI.
Account Hijacking and Internal Network Attacks in Kubeflow
Exploiting exposed or weakly authenticated Kubeflow dashboard environments to execute internal network pivot attacks.
PAIT-ARV-100 (Archive Slip Bugs in ML Models)
Zip Slip and Tar Slip vulnerabilities in model archive formats (like Model-Archive or PyTorch .pth zips) overwriting critical system files.