Risk & Regulation

TechWolf's system infers skills and provides insights, both for individuals and other entities like jobs and courses. The main focus of our efforts around fairness, risk, and regulation is the application of our technology to employees, as our data and insights can be used to inform people-impacting decisions. We proactively assess and mitigate the risks tied to the use of AI with people data.

Bias Audits

TechWolf performs internal bias audits for all major model updates and makes the audit reports available to customers. Within these audits, we use a real-world dataset to detect harmful bias against several protected attributes. We use disparate impact analysis, a technique for which 80% is widely considered the minimum score at which a system is safe and unbiased enough for use. In all tests, TechWolf’s AI achieves scores over 90%, greatly outperforming this baseline.

In addition to our own testing, we also provide customers with a bias testing toolbox, which enables them to perform the same evaluation on their own data.

Risk Framework

Our development process is covered by our ISO27001 certification, which includes explicit management of risks. In addition to this, we leverage an internal AI risk framework, which is aligned with the ISO/IEC TR 24027:2021 bias standard. AI risk is evaluated on a quarterly basis as well as before and after significant product or engineering initiatives.

Privacy & Security

Our broad approach to privacy and security is described in the security section. In addition to the general measures we take for the development and deployment of our product, we also take action specifically for our AI models. This includes adversarial testing of our models, as well as anomaly detection on the performance and outputs of each AI model. In specific instances, we conduct red teaming or an external penetration test for more sensitive models.

Regulation

We track active and upcoming AI regulations, including the New York AI law and EU AI Act, and proactively ensure our technology is compliant. Customers and prospects receive access to detailed documentation around legal requirements and compliance, as well as being supported by our AI & compliance teams where needed. We maximally align our approach with leading standards such as ISO/IEC TR 24027:2021, and plan to pursue certification when sufficiently mature options become available.