AI systems and agent security
Map trust boundaries across models, tools, data and users; discuss prompt injection, permissions and human approval.
Start with your business context and systems, clarify workflow bottlenecks, AI opportunities and security requirements, and agree on practical priorities.
Map trust boundaries across models, tools, data and users; discuss prompt injection, permissions and human approval.
Review identity, access paths and account impact to prioritise architectural improvements.
Examine how data, detection methods and workflows fit together, including false alerts, evidence and analyst decisions.
Review policies, controls and evidence to clarify security governance and ISO 27001 preparation work.
Review repetitive tasks, data and tools to identify use cases and pilot priorities.
Agree on evaluation measures, human review responsibilities and procedures, with training to help teams take over.
Project-based or ongoing advice can be discussed according to the problem and the team’s needs.
Agree on the business context, decision-makers, deployment stage and outcomes.
Review architecture, data flows, permissions and existing controls within the agreed scope.
Connect risks to controls and discuss feasibility, responsibilities and priorities.
Deliver the agreed analysis and recommendations, with training or follow-up advice where appropriate.
Collaboration with Billows Tech. on deception technology and with Zyell Solutions on SEDE anomaly detection connects research methods with industry problems.
Research on agent authorisation, continuous trust evaluation and AD account risk informs consulting discussions, within each study’s scope.
Explore the researchCollaboration with Askey includes an AI-enabled transformation project focused on improving business processes through AI adoption.
Explore engagement optionsSend a non-confidential outline of your context, problem and desired outcome. We can then agree on the scope, inputs, timeline and deliverables.
Discuss consultingYes. Start with the workflow you want to improve, your data and team context, then discuss tool selection, pilot scope and security requirements.
Deliverables may include a risk inventory, control recommendations, prioritised improvements, architecture review notes or training. Scope, required inputs and acceptance criteria are agreed before work begins.
Start with a non-confidential outline of the business context, deployment stage, current problem and desired outcome. We can then discuss scope, timing and the engagement format.
Wilbur Wei consults on zero trust and AD risk, reviewing identity, access paths and account impact to prioritise architectural improvements. At SEMICON Taiwan 2025 he spoke on “Silicon Shield Reinforcement: AD Security Strategies and Multi-Factor Defense in the Semiconductor Industry”. His research includes the STIE-ZTA trust inference framework for zero trust (CISC 2025 Honorable Mention) and an Active Directory high-privilege account risk model. Send a short non-confidential outline of your business context to wilbur.wei.cyberai@gmail.com.
Four steps: context and scope, review and assessment, recommendations and priorities, and deliverables and follow-through. Engagements can be project-based or ongoing, depending on the problem and the team’s needs; scope, required inputs and acceptance criteria are agreed before work begins.
Share your event or team context, the problem to address and the expected timeline.
Prepare an inquiry, then copy it or choose an email option to send it.