Cyberis Blog
Reassuringly clear thinking.
- Penetration testing
- Red teaming
- Research
- Tools and techniques
Probabilistic Systems, Deterministic Security
LLMs and AI agents are increasingly being connected to tools, APIs, data sources and business workflows. While this creates real value, it also introduces an important security question: should probabilistic systems be trusted to enforce security boundaries? This article explains why prompts, refusals and LLM-based guardrails should not be treated as access controls. It explores the difference between deterministic security enforcement and probabilistic model behaviour, highlights risks seen in real-world AI agent testing, and sets out why secure LLM architectures should keep deterministic controls in charge of identity, authorisation, data access, tool execution and sensitive output handling. The key principle is simple: use LLMs at the interaction layer, but enforce security in the surrounding application, services and infrastructure.
- Penetration testing
- Research
- Tools and techniques
Keeping the Consultant in the Loop: How AI Supports Our Security Testing Strategy
Artificial intelligence is rapidly changing the security testing landscape, but effective penetration testing still depends on experienced consultants exercising judgement, creativity, and accountability. In this blog, we explain how Cyberis is using AI to accelerate analysis, improve efficiency, and support consultants during engagements, while ensuring that every finding remains validated and consultant-led. We also explore the risks, limitations, and security considerations around AI usage, including data handling, model deployment, and why human expertise remains critical when assessing real-world business risk.
- Penetration testing
- Research
- Tools and techniques
When project files become instructions: AI agents, CI pipelines and the new attack surface
AI agents now read repository files, skills, plugins and CI context as part of normal operation, which creates new attack paths across local development and automated workflows. This blog explains how those instruction channels work, why they matter from a security perspective, and what organisations should do to manage the risk before unsafe patterns become normalised.
- Penetration testing
- Tools and techniques
Common AI Implementation Mistakes to Avoid Part 5
Over the course of four articles, we have explored some of the most common AI implementation mistakes which have been observed during real tests. In this final article, we bring them together and outline the potentially devastating consequences for a business.
- Penetration testing
- Tools and techniques
Common AI Implementation Mistakes to Avoid Part 4
With the exploding use of AI in internal and external applications being rapidly deployed across all sectors, new chatbots, AIs, and versions are being constantly released. Whilst this offers exciting new capabilities, it also widens the potential attack surface of a company’s infrastructure – particularly if updates are not duly applied.
- Penetration testing
- Tools and techniques
Common AI Implementation Mistakes to Avoid Part 3
The use of AI in internal and external applications is rapidly being deployed across all sectors with great alacrity. Whilst this offers exciting new capabilities, it also widens the potential attack surface of a company’s infrastructure. AI chatbots, even with filters, can be tricked into breaking ethical barriers, which could lead to serious consequences.
- Penetration testing
- Tools and techniques
Common AI Implementation Mistakes to Avoid Part 2
The use of AI in internal and external applications is rapidly being deployed across all sectors, often handling vast arrays of data from across an organisation. Whilst this offers exciting new capabilities, it also widens the potential attack surface of a company’s infrastructure. AI infrastructure can accidentally expose sensitive data to unauthenticated users if datasets are not properly configured.
- Penetration testing
- Tools and techniques
Common AI Implementation Mistakes to Avoid Part 1
The use of AI in internal and external applications is rapidly being deployed across all sectors. Whilst this offers exciting new capabilities, it also widens the potential attack surface of a company’s infrastructure. AI chatbots can accidentally expose sensitive data if permissions aren’t properly configured.
Improve your security
Our experienced team will identify and address your most critical information security concerns.