AI compliance policy

1. Transparency and user awareness

1.1. Disclosure of AI-generated content

All AI-generated content and responses will be clearly disclosed to users.

Transparency is a central principle of our approach to AI-powered customer service. Users should be aware when content or responses have been generated by artificial intelligence and should be able to understand the role of AI in the interaction.

The purpose of disclosure is to make the involvement of AI clear. The ability of an AI system to generate natural language responses does not change our commitment to identifying those responses as AI-generated.

This commitment concerns the presentation of AI-generated content to users and forms part of our broader approach to transparent customer interactions.

1.2. Identification of AI-assisted content

Any AI-assisted interactions that modify or generate content will be labeled accordingly.

This includes identifying the involvement of AI where content is generated or modified through an AI-assisted interaction. The purpose of labeling is to make the origin or modification of that content apparent to the user.

The same transparency principle applies to the generated media described in Section 4 of this policy. Whether the output is a response or another form of generated content, our commitment is to disclose the involvement of AI clearly.

1.3. No cognitive behavioral manipulation

Komsai AI does not manipulate user behavior or decision-making through cognitive behavioral techniques.

Our use of AI in customer service is governed by this commitment alongside the requirements for transparency and user awareness. The generation of conversational responses does not alter our position on the use of cognitive behavioral techniques to manipulate users.

Disclosure and non-manipulation address different aspects of responsible interaction: disclosure concerns awareness of AI involvement, while non-manipulation concerns the techniques used in the interaction.

2. Ethical AI and risk management

2.1. Excluded activities

Our AI systems do not engage in biometric identification, social scoring, or real-time surveillance.

These exclusions form part of the boundaries described in this policy for Komsai AI. They apply alongside our commitment to developing customer service agents that are safe, transparent, traceable, and non-discriminatory.

The use of AI to support customer interactions does not change these stated exclusions. Biometric identification, social scoring, and real-time surveillance are not activities carried out by our AI systems.

2.2. Sensitive biometric data

Komsai AI does not process or store sensitive biometric data.

This commitment concerns both the processing and the storage of such data. It complements the exclusion of biometric identification described above and is a separate statement from our general commitments regarding personal data protection.

Our position on sensitive biometric data should therefore be read together with, rather than as a replacement for, the data protection provisions in Section 5.

2.3. Continuous assessment of unacceptable risks

Our models are continuously assessed to ensure they do not pose unacceptable risks as defined by the AI Act.

Risk assessment is an ongoing element of our approach to responsible AI. It concerns the risks associated with the models and their use, rather than only the ability of a system to produce a response.

The assessment of unacceptable risks sits alongside the ongoing impact assessments described in Section 3. Together, these commitments address both the boundaries established by the AI Act and the fairness, accuracy, and safety of customer interactions.

Our commitment to ethical AI combines these risk assessments with the explicit exclusions set out in this section.

3. Compliance with high-risk AI regulations

3.1. Position regarding high-risk categories

Komsai AI does not fall under high-risk categories, such as law enforcement, migration control, or medical devices.

This is the position stated in this policy for our AI solutions. The examples above identify the types of high-risk categories referred to in the original scope of this commitment.

Alongside this position, Komsai voluntarily upholds best practices in AI governance. Our approach includes ongoing impact assessments, human oversight, and the availability of human review for complaints about AI decisions.

3.2. Ongoing impact assessments

We undertake ongoing impact assessments to ensure fairness, accuracy, and safety.

These three considerations address different aspects of AI performance in customer service. Fairness concerns the non-discriminatory approach described in this policy. Accuracy concerns the quality and correctness of AI outputs. Safety concerns the potential for harmful outcomes in customer interactions.

The commitment to ongoing assessment means that these considerations remain part of our approach beyond the initial development of an AI solution.

Impact assessments complement our continuous evaluation of AI performance and our assessment of unacceptable risks. They form part of the governance practices that Komsai voluntarily upholds.

3.3. Human oversight

We maintain human oversight to prevent harmful automation in customer interactions.

Human oversight forms part of our approach to the responsible use of AI-powered customer service agents. Its purpose is to address the risks associated with automation in the context of interactions with users.

This commitment is connected to the fairness, accuracy, and safety considerations described above. It also complements the availability of human review when users raise complaints about AI decisions.

3.4. Complaints and human review

Users have the right to file complaints about AI decisions, with human review available.

The availability of human review provides a human perspective when a user raises a concern about an AI decision. This forms part of the governance commitments described in this section.

Human oversight and human review serve related purposes. Oversight concerns the prevention of harmful automation, while review concerns the examination of a complaint about an AI decision.

Both are included in Komsai’s approach to responsible customer interactions.

4. Generative AI and content responsibility

4.1. Design to prevent illegal or harmful content

Our AI models are designed to prevent the generation of illegal or harmful content.

Content responsibility forms part of the design of our generative AI solutions. The ability to generate content is considered alongside the safety and ethical principles set out in this policy.

This commitment concerns the design of the models to prevent illegal or harmful outputs. It is connected to the ongoing assessment of risks and the continuous evaluation of AI performance described elsewhere in this policy.

The generation of content, its responsible design, and its transparent presentation are related elements of our approach.

4.2. Copyright compliance

We comply with EU copyright laws and publish summaries of any copyrighted data used for AI training.

Our copyright commitment addresses the use of copyrighted material in the training of AI models. The publication of summaries forms part of the transparency described in this policy in relation to that training data.

This commitment is distinct from labeling AI-generated outputs. Training data summaries concern copyrighted data used for training, while output labeling concerns the content presented to users.

Both provisions form part of our approach to responsibility and transparency in generative AI.

4.3. Labeling of generated images, videos, and audio

Any AI-generated images, videos, or audio will be clearly labeled to avoid misinformation.

This commitment extends the disclosure principle beyond conversational responses to generated media. The purpose is to make the AI-generated nature of the material clear.

The format of an output does not change the transparency principle. Images, videos, and audio generated by AI are subject to the stated commitment to clear labeling.

Our approach therefore combines the design of models to prevent illegal or harmful content with the identification of generated content when it is presented.

5. Data protection and security

5.1. Compliance with data protection requirements

Komsai fully complies with the General Data Protection Regulation (GDPR) and data privacy laws across all markets in which we operate.

Data protection is part of our commitment to responsible AI development and deployment. The use of AI in customer service is considered alongside the requirements governing personal information.

This commitment applies together with the specific provisions below concerning personal data retention, encryption, and cybersecurity.

Data protection and AI governance address related responsibilities. The commitments in this section concern the protection of user information, while the other sections address matters such as transparency, risk assessment, generated content, and human oversight.

5.2. Explicit consent for personal data retention

No personal data is retained without explicit user consent.

This is our stated commitment regarding the retention of personal information in the context of the AI interactions covered by this policy.

The consent requirement concerns personal data retention and should be read alongside our broader commitment to compliance with GDPR and applicable data privacy laws.

It also sits alongside the separate provision that Komsai AI does not process or store sensitive biometric data.

5.3. Encryption of AI interactions

AI interactions are encrypted to protect user information.

Encryption forms part of the security commitment described in this policy. It addresses the protection of information in AI interactions and complements our requirements concerning data privacy.

The use of AI-generated responses does not alter this commitment to the protection of user information.

5.4. Cybersecurity standards

AI interactions follow strict cybersecurity standards to protect user information.

Cybersecurity and encryption are complementary elements of the security approach stated in this policy. Together, they support the protection of information involved in AI-powered customer interactions.

These security commitments form part of our overall approach to responsible AI, alongside transparency, ethical risk management, and human oversight.

6. Continuous monitoring and reporting

6.1. Continuous evaluation of AI performance

Komsai continuously evaluates AI performance.

This evaluation forms part of our ongoing approach to the responsible use of AI solutions. It is connected to the commitments concerning fairness, accuracy, safety, and the prevention of illegal or harmful content.

Performance evaluation should be read alongside the continuous assessment of unacceptable risks and the ongoing impact assessments described in this policy.

These commitments address related aspects of AI use: performance, impact, and risk. Together, they form part of our approach to monitoring AI-powered customer service.

6.2. Reporting of serious incidents

Komsai reports any serious incidents as required by the EU AI Act.

This commitment concerns the reporting requirements applicable under the AI Act. It forms part of our approach to regulatory compliance and responsible deployment.

Serious incident reporting is distinct from the availability of human review for user complaints. The former concerns reporting as required by the AI Act; the latter concerns the review of complaints about AI decisions.

Both commitments are included in this policy and should be understood according to their respective purposes.

6.3. Cooperation on compliance and ethical standards

We work with regulatory bodies and industry partners to ensure our AI remains compliant with evolving laws and ethical standards.

This cooperation forms part of our approach to continuous compliance. It reflects the relationship between the development of AI solutions and the legal and ethical standards relevant to their use.

Our commitment includes attention to evolving requirements, alongside the continuous evaluation and risk assessment described in this policy.

Regulatory compliance, ethical standards, and performance monitoring remain connected elements of our approach to responsible AI.

7. Supporting innovation and responsible AI

7.1. Openness to EU AI testing environments

Komsai is open to participating in EU AI testing environments to refine our models before deployment.

This statement expresses our openness to these environments as a means of supporting model refinement in advance of deployment.

The opportunity to refine AI models is connected to the principles described throughout this policy, including safety, fairness, accuracy, and responsible development.

Our openness to participation forms part of our approach to innovation within the framework of responsible AI.

7.2. Fair and accessible adoption

Komsai promotes fair and accessible AI adoption.

This commitment concerns the adoption of AI by businesses and is connected to our broader approach to responsible development and deployment.

We consider responsible innovation alongside the ability of businesses to benefit from AI-powered technology. Transparency, ethical use, data protection, and human oversight remain part of that approach.

7.3. Businesses of different sizes

Komsai promotes the adoption of AI technology that benefits businesses of all sizes, including small and medium-sized enterprises and larger enterprises.

Our approach to innovation includes both these groups. The principles described in this policy apply to our commitment to responsible AI adoption across businesses of different sizes.

Supporting innovation and upholding responsible AI principles are connected parts of Komsai’s approach. Our commitment is to develop and deploy AI solutions within the framework of transparency, safety, non-discrimination, data protection, and compliance described in this policy.