Issued by Talkative Ltd in connection with the EU AI Act (Regulation (EU) 2024/1689)
This statement is provided for informational purposes to support distributor and deployer transparency obligations under the EU AI Act. It is not a warranty, guarantee, or certification of compliance.
- Issuer Talkative Ltd (company no. 09660808), Wesley Clover Innovation Centre, Newport NP18 2YB, United Kingdom ("Talkative") is the provider of the AI systems described in this statement, which covers the Talkative AI product line as designed and supplied as at the date of issue, including chat AI, voice AI, and agent-assist features.
- Roles under the EU AI Act Talkative is the provider of the AI system. Customers operating the system to interact with their own end-users are the deployers. Resellers of the Talkative product are distributors.
- Intended purpose and risk classification The Talkative AI product line provides conversational AI capabilities — chat, voice, and agent-assist — supplied to support customer service and contact-centre workflows. It is not designed for biometric identification or categorisation, emotion recognition from biometric data, workplace monitoring, eligibility decision-making, or any other use case classified as high-risk under Annex III or prohibited under Article 5 of the Act. The product line sits outside the Annex III high-risk categories, and the primary applicable obligation is end-user transparency under Article 50.
- End-user transparency (Article 50) The product is designed to support deployer compliance with Article 50. Deployers configure end-user-facing messaging, including any disclosure that an end-user is interacting with an AI system. Talkative's customer success function provides guidance and strongly recommends that an AI disclosure is implemented as part of every deployment.
- Foundation models and sub-processors The Talkative product integrates third-party foundation models from established providers. Foundation model obligations under the Act's General Purpose AI regime sit with those upstream providers. A current list of sub-processors and model providers is maintained by Talkative and made available to customers under the existing data processing agreement.
- Use of customer data for model training and improvement Talkative does not train any proprietary AI model on customer data, and does not operate a proprietary large language model in production. Customer data is processed only for the purposes of delivering the Services, as set out in the Data Processor Agreement (DPA). Where third-party foundation models or AI services are integrated into the Talkative product, the data handling and training-data policies of those providers apply; the providers and their published policies are listed in Sub-appendix B of the DPA.
- Sentiment indicators The product may include features that analyse the text content of customer interactions to provide sentiment indicators supporting routing and agent prioritisation. These features process interaction content for customer service purposes and are not designed for biometric emotion recognition, workplace monitoring, or any use case within Annex III or Article 5 of the Act. Deployers are responsible for ensuring their use remains within this intended purpose.
- Deployer responsibilities Several obligations sit with the deployer, including: defining the specific use case; informing end-users they are interacting with AI; managing knowledge base content; carrying out any required data protection impact assessment; and ensuring the deployment does not fall within the high-risk or prohibited categories of the Act.
- Limitations This statement reflects the Talkative product line as designed and supplied as at the date of issue. Talkative's products and the regulatory framework will continue to evolve. The statement is provided for information only and does not constitute a warranty, guarantee, or certification of compliance with the EU AI Act or any other regulation. Talkative does not assume responsibility for use cases, configurations, or deployments determined by the deployer.