Audits & Assessments
Ensuring compliance at every step.
"With great power comes great responsibility." This expression captures exactly the immense potential of Artificial Intelligence (AI). It drives innovation, but also has potential adverse impact on privacy, ethics, human rights, and the environment.
Rhite offers a team of experts in identifying and managing such risks, helping you build responsible solutions, from the technical details of applications to the policies at organizational level.
Do you want to request an audit or assessment of an Artificial Intelligence-based tool? Contact us or drop by for a visit.
Types of assessments
AI Impact Assessments
We can examine solutions to determine the potential impact on individuals and society. For this we use different tools and impact assessment frameworks.
Possible use cases include:
- Assessment of an AI solution during development;
- Assessment of a third-party AI solution before use;
- Due diligence of AI solutions, for investors.
We guide your organisation during the performance of different types of impact assessments such as: IAMA, ALTAI, DPIA, PLOT4ai, FRIA and more. Depending on the type of organisation you are (government, large or medium organisation, SME, start-up) and the type of solution and context, we will guide you through the impact assessment tool that is more adequate for your specific circumstances and legal requirements.
We can facilitate the sessions and the whole process from beginning till end. We can also provide custom-made training sessions for your organisation to learn how and when to implement the different assessment tools during and after the development life cycle of a solution.
AI Bias Assessments and audits
Our AI Bias Audit service offers impartial assessments of your AI systems to identify biases using advanced methods and tools adapted to your specific use case, ensuring fairness and compliance with country-specific regulations. Depending on your needs, we can review your AI model’s architecture, analyze training data, perform statistical bias detection tests, and evaluate human oversight and governance processes. Our deliverables include a comprehensive audit report with findings and bias mitigation recommendations, as well as ongoing strategies for monitoring and compliance. Our methods are always based on the latest research available.
Red teaming Large Language Models (LLMs)
We provide specialized Red Teaming services for LLLMs designed to uncover vulnerabilities and ensure both the security and safety of your AI systems. By employing advanced, use case-specific techniques, we help make sure your models are resilient against potential threats and do not cause harm to others. Our services include simulating adversarial attacks, testing the model’s response to harmful inputs, checking for copyright infringement, and evaluating existing defense mechanisms. You will receive a thorough report detailing our findings, along with actionable recommendations for mitigating risks and enhancing your AI's robustness and safety. We also offer strategies for continuous monitoring and improvement to keep your systems secure and compliant over time.
AI Audits vs Assessments
What is the difference between an assessment and an external audit?
In both cases, systems and organisations can be assessed based on different controls. These controls usually come from an approved standard or framework. Currently there is not such an approved European standard for auditing AI systems and organisations and that is why there are not accredited bodies that can offer audit certifications yet.
This doesn’t mean your organisation cannot be audited by an external party. An external party can audit your system based on available frameworks, sometimes only based on interviews and review of documentation and other times also analysing your system or having together with you a look at it.
The result of the audit will be a report indicating what is the compliance status that your system and organisation has according to the requirements and measurements of the controls assessed.
An assessment is a similar process that an audit, with the difference that the company performing the assessment can also offer guidance to the organisation to implement the right controls and increase the maturity level. An external audit should be done by an impartial party. This means that the same party offering guidance and consultancy should not be the same party auditing the system.
Privacy Impact Assessments
DPIA
It is important to start with a Data Protection Impact Assessment (DPIA) on time. We focus specifically on the technically challenging DPIAs within your organisation. In addition to the legal and technical aspects, we help you find the possible risks* and assess if they are acceptable or measures need to be taken.
*We do not only look at risks related to data protection and security, but also other risks that could arise from the design and that could have an adverse impact on the indiviual and/or society.
At Rhite we provide both audits and assessments.
For more information, contact us.