AI security & governance

Securing AI systems and governing their use: threats, controls, frameworks and regulation.

AI security & governance

Data poisoning: when attackers teach your model instead of hacking it

A model learns whatever its data says, so whoever can influence the data can influence the model. How label poisoning, backdoors and web-scale poisoning work, and where in the pipeline to stop them.

AI security & governance

Similarity is not authorisation: why RAG assistants leak, and how to stop it

Retrieval-augmented assistants find the most relevant document for a question, not the one the user is allowed to read. Why that leads to oversharing, and the controls that fix it at the retrieval layer.

AI security & governance

The EU AI Act after the Digital Omnibus: the new timeline for Asian companies

The EU's Digital Omnibus on AI took effect in July 2026 and pushed back the high-risk deadlines. What still applies now, what moved, and why companies in India and Asia should care.

AI security & governance

Prompt injection explained, and the controls that actually help

Prompt injection is the top risk for applications built on large language models, and it can't be fully patched. How direct and indirect injection work, and the layered controls that limit the damage.

AI security & governance

Securing AI agents: permissions, tools and the confused-deputy problem

AI agents don't just answer questions; they call tools, read data and take actions. Why that changes the threat model, and the identity, authorisation and monitoring controls agents need.

AI security & governance

AI risk assessment: a five-step method for security and risk teams

A practical way to assess the risks of an AI use case, from understanding the context to choosing controls and monitoring, using ideas from the NIST AI RMF and ISO/IEC 42001.

AI security & governance

ISO/IEC 42001: what an AI management system involves

ISO/IEC 42001 is the first certifiable management-system standard for AI. What it requires, how it relates to ISO 27001 and the EU AI Act, and how to get started without a big programme.

AI security & governance

Shadow AI: writing an acceptable-use policy people will follow

Staff are already using AI tools at work, approved or not. Banning them rarely works. How to write a short, practical generative AI policy and back it with approved tools and simple controls.

RSS feed · Articles are general information, not legal advice.