TaoQ AI Labs
Labs
Our research, open-source tools and writing, published so anyone can read and check them. For client work, see Work.
Open source
Tools we build and maintain, and the projects we contribute to.
Creator and maintainer
Ziran: open-source AI agent security testing
Finds vulnerabilities in AI agents: dangerous combinations of tools, side effects of what agents actually execute, and trust exploited over several steps. Apache-2.0.
CREATOR
WUMING
A zero-dependency Go library that finds and redacts personal data in text: 75+ detectors across 14 locales, with presets for GDPR, the EU AI Act, HIPAA and PCI-DSS.
CREATOR
WUWEI
A Claude Code plugin that runs a chartered team of AI agents across your repositories. Every change is reviewed by an agent that did not write it, and hooks guard the irreversible steps.
CREATOR
QKD Playground
An interactive quantum key distribution simulator. Step through the BB84, B92, E91 and SARG04 protocols on real Qiskit simulation, add an eavesdropper, and watch the error rate expose her.
CONTRIBUTOR
NVIDIA Garak
Contributions to the Garak LLM vulnerability scanner, the open-source tool used to probe LLMs for prompt injection, jailbreak susceptibility, and other failure modes.
CONTRIBUTOR
Hugging Face Transformers
Contributions to the Transformers library, a widely used open-source framework for working with large language models.
Research
Research and prototype work delivered through TaoQ AI, and our founder's published research. Some details are sanitised for client confidentiality.
Synthetic-data and temporal-detection platform for online romance fraud
AI researcher and architect leading the platform architecture, the synthetic-data pipeline, and the temporal-detection framework.
Business case
Romance-scam detection tools tend to look for malicious content, while scam messages work through emotional manipulation, so much of this fraud is hard to detect. We found no commercial training datasets for the conversation patterns that matter.
Approach
A hexagonal-architecture platform combining a configurable synthetic-conversation pipeline grounded in published criminological research on scam progression, and a temporal behavioural detection service that fuses rule-based pattern checks, statistical conversation-dynamics analysis, and ML classification trained on the synthetic corpus.
Deliverables
Vendor-neutral, multi-provider LLM stack with API-first scoring output, fully synthetic training data (no victim data), and a research-collaboration framing aimed at academic partners in fraud and LLM safety.
Evaluating the Feasibility of MITRE-Aware Large Language Models for Tactics, Techniques, and Procedures Detection in Cyber Threat Intelligence Reports
MSc Artificial Intelligence (Distinction), University of Bath. Dissertation by our founder.
Dissertation
Asks whether a domain-adapted Large Language Model fine-tuned against the MITRE ATT&CK framework can automate Tactics, Techniques, and Procedures (TTP) detection in unstructured Cyber Threat Intelligence reports, where manual parsing currently leads to inconsistency and missed threats. Combines causal-language-model fine-tuning over a transformer baseline with a two-stage training strategy and a hybrid evaluation regime (ROUGE plus LLM-as-judge qualitative analysis). Finds that domain-specific fine-tuning improves precision on shorter text segments but trades coherence on longer outputs and produces mixed results when mapping to specific ATT&CK techniques, pointing to hybrid approaches with knowledge graphs and continual learning as the more promising path.
Writing
Articles on AI architecture, security and regulation.
- EU AI Act: What Your Engineering Team Actually Needs to Change
- Your AI Agent Has a Backdoor, Here's How I Found It
- Lessons from building AI architecture at scale
Research collaboration enquiries
Open to academic partners, advisory boards, and joint research on synthetic data, AI agent security testing, and engineering for AI regulation.
Email us