As an internationally oriented centre of excellence, acib GmbH welcomes the enormous potential that artificial intelligence offers for industrial biotechnology, research and administrative processes. We do not view this rapid technological development as a threat, but rather as a fundamental catalyst that enhances our innovative strength and enables new scientific paradigms. AI technologies allow us to penetrate complex data landscapes more efficiently and rethink established research processes. We expressly encourage our employees to be open to this technology, to actively integrate it into their everyday work and to boldly seize the opportunities it offers.
acib always views the use of AI as a social and ethical responsibility: we are guided by the principles of good scientific practice, the provisions of the EU AI Act and generally accepted ethical guidelines (fairness, non-discrimination, transparency, data protection and environmental protection). On this basis, acib follows a number of important guidelines:
- AI as a tool and intellectual sparring partner: For acib, intelligence – both human and artificial – is the most important factor of production. We consistently regard AI as a powerful tool and intellectual sparring partner, but never as an end in itself or a substitute for human expertise. Whether as a ‘co-author’ in structuring texts, brainstorming new research approaches or identifying hidden patterns in large data sets, AI serves as a source of inspiration that challenges our researchers and significantly speeds up routine tasks. This creates valuable freedom for our teams to focus even more intensively on highly complex, creative and value-adding activities.
- Human responsibility and ‘human in the loop’: We firmly believe that humans, with their experience, creativity and unique judgement, are at the heart of scientific progress. At acib, the use of AI is therefore always complementary. The principle of human oversight (‘human in the loop’) applies without restriction: AI systems generate suggestions, texts and analyses, but the final interpretation and decision-making are always the responsibility of our experts. Automated conclusions or unreflective adoption of AI results without critical human review are strictly excluded in order to avoid scientific misinterpretations.
- Scientific integrity and trust: An open, supportive approach to artificial intelligence also requires a high degree of responsibility. Maintaining scientific integrity, objectivity and public trust in our research work is our top priority. We require that the results of AI systems be scrutinised with the same critical distance and scientific rigour that is applied in academic discourse with human colleagues. Transparent handling of AI-generated content, protection of sensitive research data and awareness of possible algorithmic bias form the non-negotiable foundation of our AI culture.
The use of AI systems is only permitted in accordance with the approval levels defined by acib, which range from safe systems (AI-WHITE) to strictly prohibited applications (AI-RED). High-risk systems according to the European AI Act require explicit prior approval by management and a detailed risk assessment. The development of proprietary AI models is strictly tied to approved research projects and must be planned in detail in advance.
Limits on the use of AI and absolute prohibitions
Even though acib GmbH actively promotes the potential of AI, there are clear, non-negotiable limits on its use. These serve to protect scientific integrity, comply with the European AI Act and uphold fundamental ethical standards.
The use of AI systems at acib is strictly prohibited in the following areas and activities:
- No automated final decisions: AI systems must never make final decisions unsupervised, especially in areas where incorrect interpretations can lead to scientific misjudgements. The final say and responsibility for content must lie with a human expert (‘human in the loop’).
Prohibition of use for unethical purposes: AI results that promote discrimination, contain bias or contradict the organisational values and ethical standards of acib GmbH may not be adopted. - Exclusion of high-risk and prohibited systems without approval: The use of AI systems classified as high-risk under the AI Act is strictly prohibited without explicit prior approval from management and a prior risk assessment. Similarly, unauthorised systems declared internally as ‘AI-RED’ may not be used for business purposes under any circumstances.
- Entering sensitive data into unsecured systems: It is strictly prohibited to enter confidential project information or personal data into public, non-explicitly secured (AI-WHITE) or untested AI systems. This applies in particular to data that is not intended for public consumption, as public systems cannot guarantee sufficient protection of confidentiality.
- Avoiding unreflective information transfer: At the beginning of 2026, language models and generative AIs still tend to ‘hallucinate’ false quotations or facts in some areas. It is prohibited to consider AI-generated references, source citations or scientific claims as verified and to include them in publications without first personally verifying their authenticity and relevance against the original sources.
These strict guidelines are in no way intended to limit the innovative power of our teams, but rather to create a safe, responsible and legally compliant framework for our research. Consistent adherence to these boundaries is essential to protect acib GmbH and all employees from legal risks and to maintain our status as a trustworthy, scientifically sound, leading research institution in the long term.
Responsible science in the age of AI
Public trust in the research conducted by acib GmbH is based on absolute transparency and scientific integrity. The following guidelines therefore apply to all publications, reports, funding applications and external communications relating to AI systems:
- Clear authorship and responsibility: Scientific publications, official communications and project reports are written by humans. Artificial intelligence can be used as a tool to support the writing or analysis process, but can never act as an author in its own right. Researchers always bear full responsibility for the accuracy, originality (including plagiarism checks) and scientific validity of published results. This means that the copyrights to original works always remain in the hands of the researchers, or the full rights of use are retained by acib in accordance with copyright law and service contracts.
- Mandatory labelling: As soon as AI-generated content (texts, images, diagrams or simulated data) is presented or published externally, it must be declared transparently and uniformly. Depending on the extent of AI involvement, designations such as ‘AI-GENERATED’ (for content created entirely by AI) or ‘GENERATED WITH AI SUPPORT’ (for AI-supported elaborations) must be used. In scientific publications, the significant use of AI – for example, in complex data analyses – must also be explained in detail in the methods section or in the acknowledgements.
- Exceptions to the labelling requirement: To avoid ‘over-declaration’, explicit labelling is not required if AI tools are used solely for purely supportive routine tasks that do not alter the scientific content. This includes, for example, automatic spelling and grammar corrections, formatting or AI-supported translations, provided that these have been verified by human expertise.
- Scientific reproducibility and documentation: The use of AI-supported evaluations must not compromise the traceability of research results. All AI-supported analyses must be documented (e.g. by storing prompts and parameters in laboratory protocols) in such a way that external reviewers or peers can understand the process and reproduce the results. This applies in particular to data-driven AI methods and machine learning.
- Transparency in project planning and funding applications: If artificial intelligence is used specifically in research projects, this must be clearly stated in the project planning, including a corresponding risk assessment. Similarly, when preparing national, European or international funding applications, it must be checked and disclosed in advance to what extent AI systems
With these transparency and documentation requirements, we underline our commitment to excellent, comprehensible and trustworthy research. The clear declaration of AI use, the complete reproducibility of our methods and the inviolable human responsibility for publications ensure that acib GmbH remains a guarantor of the highest scientific integrity even in the age of artificial intelligence.
Data protection and confidentiality
When using AI systems, the protection of confidential information, internal company data and personal data has the highest priority. The processing of sensitive data is strictly prohibited, unless it involves specially secured and expressly approved AI systems (category AI-WHITE).
Microsoft 365 Copilot as a secure standard solution
To enable our employees to use AI-supported assistance functions securely in their everyday work, acib GmbH provides Microsoft 365 Copilot as an officially approved and integrated tool. As this system operates within our secure corporate environment (‘tenant’), special protection mechanisms apply:
- Data entered into or processed by Microsoft 365 Copilot (prompts, document content, emails, chats) always remains within the acib environment.
- This data is not used by Microsoft for training public language models (foundation models).
- Copilot respects the existing access rights and sensitivity labels (Information Rights Management/Sensitivity Labels) of acib GmbH. It only accesses data for which the respective user already has read permission.
AI tool directory and system evaluation
In addition to Microsoft 365 Copilot, other AI systems, including open ones, can be used provided they have been tested and categorised in advance. acib GmbH maintains a central, binding directory of all approved AI tools on its intranet portal (data.acib).
- Five-level approval model: In this directory, each AI system is assigned a specific approval level – from AI-WHITE (secure environment for restricted data processing) to AI-RED (strict usage ban). The respective approved areas of application are also defined there.
- Applying for new systems: Employees can use data.acib to actively request the expansion of areas of application or the testing and approval of AI systems that are not yet listed. Use is only permitted after official approval has been granted.
Personal responsibility for data entry
Despite the use of secure systems such as M365 Copilot, users remain personally responsible. The following principles must be observed before using AI tools:
- No public tools for protected data: Entering confidential project information, research data prior to publication, or personal data into public AI services (such as freely accessible web versions of ChatGPT, Claude, etc.) is strictly prohibited, as confidentiality cannot be guaranteed.
- Review of non-disclosure agreements (NDAs): Before use, users must check on their own responsibility whether the processing of specific project information by AI systems is in accordance with existing contractual confidentiality requirements (non-disclosure agreements) with external partners.
- Data minimisation: The principle of data minimisation (‘need-to-know principle’) should always be applied. Only provide AI systems with the context that is absolutely necessary for the fulfilment of the respective task.
AI competence and training
The rapid development in the field of artificial intelligence requires not only powerful technologies, but above all competent and critically thinking users. In accordance with the applicable legal provisions (in particular the European AI Act) and our internal quality standards, the professional use of AI systems at acib GmbH is subject to clear qualification requirements
- Mandatory training as a basis: The use of AI systems for business purposes is only permitted for employees who have been specially trained in the use of AI and can demonstrate the necessary expertise. Although we expect our researchers to proactively follow developments in their field, this does not replace the formal training requirements. All employees are therefore required to complete the mandatory AI training courses assigned to them within the specified time frame.
- Consequences of not having the required qualifications: The legal framework for the use of AI is strict. Failure to complete the mandatory training courses within the specified time frame will result in the immediate revocation of permission to use any AI systems for work purposes. The unauthorised use of AI systems without the appropriate qualification or approval constitutes serious misconduct and may result in disciplinary action.
- Optional training opportunities and personal responsibility: In order to fully exploit the innovation potential of acib GmbH, we regularly offer optional, in-depth training courses and workshops, which we strongly recommend you attend. Before using a new AI system in a professional context, it is also the responsibility of employees to familiarise themselves with the specific operating instructions, the manufacturer’s intended use and any potential limitations of the system.
- Promotion of a critical AI culture: At the heart of our training efforts is the development of ‘AI literacy’. Employees are trained to always critically question the results of AI systems – which are often presented with a high degree of linguistic persuasiveness – to check for hallucinations and bias, and to validate results using alternative methods or human expertise.
The principles, limitations and guidelines listed here represent a summary of the most important guidelines for daily research and work operations. The official ‘Guideline on the use of Artificial Intelligence’ (in its currently valid version) is binding for all employees of acib GmbH. This guideline comprehensively and detailedly regulates the use of artificial intelligence in all areas of acib – from data protection-compliant use and correct documentation to strict requirements for the development of proprietary AI models in research projects. In view of the dynamic technological and legal developments, this AI guideline is intended to be a ‘living guideline’. It is regularly reviewed and further developed in order to take into account new fields of application, regulatory requirements and practical experience in a timely manner.