AI Services Terms

1. Purpose of These Terms

These AI Services Terms ("AI Terms") supplement and are incorporated into our general Terms of Service. They apply specifically to any Services involving artificial intelligence, including but not limited to AI automation, AI agents, AI consulting, AI workflow design, Retrieval-Augmented Generation (RAG) systems, AI integrations, large language model (LLM) applications, AI chatbots, and AI-driven business process automation (collectively, "AI Services"). Where these AI Terms conflict with our general Terms of Service on matters specific to AI Services, these AI Terms will govern.

By engaging DREOS STUDIO LLC ("DREOS STUDIO," "we," "us," or "our") for any AI Services, the Client acknowledges and agrees to the terms set out below.

2. Nature of AI Services

DREOS STUDIO designs, builds, integrates, and deploys AI-powered systems for business clients. This may include, depending on the project:

• Custom AI agents that perform tasks, make decisions, or take actions within defined parameters.
• Chatbots and conversational interfaces powered by large language models.
• Retrieval-Augmented Generation (RAG) systems that combine language models with Client-specific data sources.
• Workflow automations that use AI to process, summarize, classify, or act on information.
• Integrations connecting AI models (whether developed by DREOS STUDIO or third-party providers) into existing Client software and business processes.
• AI-assisted internal tooling, such as dashboards, reporting systems, or decision-support tools that incorporate AI-generated analysis.

The specific architecture, model providers, and capabilities of an AI Service are defined in the applicable Project Agreement, and DREOS STUDIO will work with the Client during the discovery phase to document the intended scope, data sources, and use cases for the AI system being built.

3. Reliance on Third-Party AI Providers

Most AI Services built by DREOS STUDIO rely, in whole or in part, on foundation models and infrastructure provided by third-party AI providers (such as large language model providers, cloud AI platforms, or embedding and vector database providers). The Client acknowledges that:

• DREOS STUDIO does not own, train from scratch, or control the underlying behavior of third-party foundation models used within a project, unless expressly agreed otherwise in a Project Agreement.
• The availability, pricing, performance, output quality, and policies of third-party AI providers may change at any time, and such changes may affect the behavior, cost, or availability of the AI Service.
• Where a third-party AI provider deprecates a model, changes its API, or discontinues service, DREOS STUDIO will make commercially reasonable efforts to notify the Client and propose a path forward (such as migrating to an alternative model), which may involve additional fees depending on the scope of work required.
• DREOS STUDIO is not responsible for outages, rate limits, cost increases, content policy changes, or performance degradation caused directly by a third-party AI provider.
• Ongoing usage-based fees charged by third-party AI providers (for example, per-token or per-request charges) are typically the Client's responsibility to budget for as an operating cost, separate from DREOS STUDIO's development or maintenance fees, unless otherwise agreed in the Project Agreement.

4. AI Outputs Are Probabilistic, Not Deterministic

The Client acknowledges that AI systems, including large language models, generate outputs using probabilistic methods rather than fixed, rule-based logic. As a result:

• AI-generated outputs may be inaccurate, incomplete, outdated, biased, or entirely fabricated — a phenomenon commonly known as "hallucination."
• The same input may produce different outputs at different times, even when using the same underlying model and configuration.
• AI systems may reflect limitations or biases present in their training data, which DREOS STUDIO does not control when using third-party foundation models.
• AI behavior may drift or change over time, including as a result of updates made by the underlying model provider, without any change made by DREOS STUDIO.
• Performance may vary across languages, dialects, domains, or edge cases not well represented in the model's training data.

DREOS STUDIO applies engineering best practices — such as prompt design, retrieval grounding, guardrails, output validation, and testing — to improve reliability where appropriate and within the agreed project scope. However, we cannot and do not guarantee that any AI system will be free from the limitations described in this Section.

5. Client Responsibility for Review and Validation

The Client is solely responsible for reviewing, testing, and validating AI-generated outputs before relying on them for any business-critical, financial, legal, medical, safety-related, or otherwise high-stakes purpose. This includes, without limitation:

• Reviewing AI-generated content, decisions, classifications, or recommendations before they are published, acted upon, or relied upon.
• Implementing appropriate human oversight ("human-in-the-loop") for AI systems that make or influence decisions with material consequences.
• Ensuring that any use of AI Services complies with applicable laws and regulations relevant to the Client's industry, including any sector-specific requirements (e.g., financial services, healthcare, employment).
• Establishing internal policies governing acceptable use of AI systems by the Client's own employees, customers, or end users.
• Communicating any regulatory or compliance context relevant to the AI system (such as consumer protection disclosure requirements) at the outset of the project, so it can be factored into the design.

DREOS STUDIO builds systems according to agreed specifications but does not accept responsibility for business decisions the Client makes based on AI-generated outputs.

6. No Guarantee of Specific Outcomes

Unless expressly stated in a Project Agreement, DREOS STUDIO does not guarantee specific quantitative outcomes from AI Services, such as a particular accuracy rate, cost savings figure, conversion rate, or automation success rate. Where performance benchmarks are agreed upon for a specific project (e.g., a target response accuracy during a defined testing phase), those benchmarks will be documented in the applicable Project Agreement and are subject to the testing conditions described therein.

7. Data Used in AI Systems

7.1 Client-Provided Data

Where an AI Service involves processing the Client's own business data (such as in a RAG system built on the Client's documents), the Client represents and warrants that it has the necessary rights to provide that data to DREOS STUDIO and to third-party AI providers for processing in connection with the Service.

7.2 Data Sent to Third-Party AI Providers

The Client acknowledges that data submitted to AI systems built by DREOS STUDIO may be transmitted to third-party AI model providers for processing, subject to those providers' own data handling and retention policies. DREOS STUDIO will document, where known and where reasonably practicable, which third-party AI providers are used in a given project so the Client can assess this against its own data governance requirements. Clients with specific data residency, retention, or training-opt-out requirements should raise these requirements at the outset of a project so that appropriate providers and configurations can be selected.

7.3 Confidentiality of Client Data in AI Systems

Data provided by the Client for use within an AI system is treated as confidential information in accordance with our Confidentiality Policy. DREOS STUDIO does not use Client-provided data to train its own general-purpose models, nor does it share Client data across unrelated Client projects, except where a third-party AI provider's own policies govern such use (in which case DREOS STUDIO will disclose this to the Client where known).

7.4 Data Accuracy and Grounding

Where a project uses Retrieval-Augmented Generation or similar grounding techniques designed to reduce hallucination by referencing Client-specific data, the accuracy of the AI system's outputs is directly dependent on the quality, completeness, and currency of the underlying data source provided by the Client. DREOS STUDIO is not responsible for inaccurate outputs that stem from outdated, incomplete, or inaccurate source data supplied by the Client.

8. AI Agents and Autonomous Actions

Where a project involves AI agents capable of taking autonomous or semi-autonomous actions (such as sending communications, modifying records, executing transactions, or interacting with other systems), the Client acknowledges the heightened importance of properly scoping permissions, guardrails, and escalation paths. DREOS STUDIO will work with the Client to define:

• The specific actions the AI agent is permitted to take.
• Escalation or human-approval checkpoints for higher-risk actions.
• Logging and monitoring mechanisms to review agent behavior.
• Rollback or override procedures in case the agent takes an unintended action.

The Client is responsible for approving the final scope of autonomous permissions granted to any AI agent prior to production deployment, and for ongoing monitoring of agent behavior once deployed, unless monitoring is separately contracted as part of an ongoing maintenance agreement.

9. Testing and Acceptance

AI Services are subject to a testing and review period as defined in the applicable Project Agreement. During this period, the Client should evaluate the AI system's behavior against agreed use cases and specifications and report any issues. Because AI systems behave probabilistically, occasional imperfect outputs during testing are expected and do not, on their own, constitute non-conformance with the Project Agreement, unless they demonstrate a systemic failure to meet the specifications agreed upon. Where a Project Agreement includes a specific, measurable acceptance benchmark (for example, an agreed accuracy threshold on a defined test set), that benchmark — rather than the presence of any individual imperfect output — will govern whether the Deliverable is deemed to conform to specification.

10. Post-Deployment Behavior Changes

The Client acknowledges that AI system behavior may change after deployment due to factors outside DREOS STUDIO's control, including updates to third-party foundation models, changes in the volume or nature of input data, or shifts in real-world conditions the system was designed to respond to. Ongoing monitoring, tuning, and adjustment of deployed AI systems is available through our monthly maintenance service and is not included by default in a one-time project engagement unless expressly stated in the Project Agreement.

11. Model and Provider Changes Over Time

DREOS STUDIO reserves the right to recommend changes to the underlying AI model or provider used in a deployed system over the life of a maintenance engagement — for example, to take advantage of improved accuracy, lower cost, or continued provider support. Any such change that would materially affect system behavior, cost, or data handling will be communicated to the Client in advance, and material changes requiring additional development work will be scoped and quoted separately unless already covered by an active maintenance agreement.

12. Human Oversight and Escalation Design

For AI Services involving customer-facing or decision-influencing interactions, DREOS STUDIO will work with the Client to design reasonable escalation paths — such as routing complex, ambiguous, or high-stakes interactions to a human team member — appropriate to the use case and agreed scope. The specific level of human oversight built into a given system depends on the risk profile of the use case and is a design decision made jointly with the Client during the scoping phase; the Client is responsible for confirming that the agreed level of oversight is appropriate for its regulatory and business context.

13. Transparency with End Users

Where an AI system built by DREOS STUDIO interacts directly with the Client's own customers or end users (such as a chatbot), the Client is responsible for determining whether applicable law or its own policies require disclosing that the end user is interacting with an AI system, and for implementing any required disclosures within the product. DREOS STUDIO can assist in implementing such disclosures as part of the agreed project scope where requested.

14. Intellectual Property in AI-Assisted Deliverables

Ownership of code, configurations, prompts, and documentation created specifically for the Client's AI Service transfers to the Client upon full payment, consistent with Section 9 of our Terms of Service and our separate Intellectual Property Policy. This does not include ownership of underlying third-party AI models, foundation model weights, or DREOS STUDIO's internal AI tooling, frameworks, or methodologies, which remain the property of their respective owners.

15. Compliance and Regulatory Considerations

AI regulation is an evolving area of law that varies by jurisdiction and industry. While DREOS STUDIO stays informed of general developments in AI governance and applies reasonable engineering safeguards, the Client is responsible for determining whether its specific use of AI Services triggers additional legal or regulatory obligations (such as disclosure requirements, algorithmic accountability rules, or sector-specific AI regulations) and for ensuring its own compliance. DREOS STUDIO does not provide legal advice, and nothing in these AI Terms should be construed as such.

16. Ethical Use

The Client agrees not to direct DREOS STUDIO to build, and DREOS STUDIO reserves the right to decline to build, AI systems intended for unlawful purposes, deceptive impersonation of real individuals, generation of misleading or harmful content, unauthorized surveillance, or other uses that DREOS STUDIO reasonably determines to be unethical or to pose significant legal or reputational risk. This includes AI systems specifically designed to evade content moderation safeguards, to generate deceptive material intended to defraud third parties, or to discriminate unlawfully against protected classes.

17. Ongoing Model Costs and Budgeting

Because many AI Services depend on usage-based third-party pricing, the Client acknowledges that operating costs for a deployed AI system (such as per-request API charges) may fluctuate based on system usage volume, model choice, and third-party pricing changes. DREOS STUDIO will provide a good-faith cost estimate during project scoping based on anticipated usage, but actual third-party costs remain the Client's responsibility and are separate from DREOS STUDIO's own fees unless otherwise agreed. Where feasible, DREOS STUDIO can implement usage monitoring or cost alerts as part of a project's scope to help the Client track ongoing third-party AI spend.

18. Decommissioning AI Systems

If the Client elects to discontinue use of an AI system built by DREOS STUDIO, the Client remains responsible for independently disabling or removing any related third-party AI provider accounts, API keys, and associated billing arrangements that remain under the Client's own ownership, in order to avoid continued third-party charges. DREOS STUDIO will provide reasonable guidance on the decommissioning process as part of an active engagement, and as a scoped, billable service after an engagement has formally ended.

19. Limitation of Liability for AI Services

In addition to the limitation of liability set out in our Terms of Service, the Client acknowledges that, given the inherent uncertainty of AI system outputs described in Section 4, DREOS STUDIO's liability for any claim arising from an AI-generated output — including inaccurate, biased, or otherwise imperfect results — is subject to the same liability cap described in our Terms of Service, and DREOS STUDIO shall not be liable for indirect or consequential damages arising from reliance on AI-generated outputs by the Client or its end users.

20. Bias and Fairness Considerations

AI models, including third-party foundation models, may reflect biases present in their underlying training data, which can manifest in outputs that are unevenly accurate or appropriate across different user groups, languages, or contexts. Where a project involves AI-driven decisions that could meaningfully affect individuals (such as screening, ranking, or classification of people), DREOS STUDIO will discuss reasonable, scope-appropriate mitigation approaches with the Client — such as testing across representative scenarios or incorporating human review checkpoints — but does not warrant that any AI system will be entirely free of bias, as this remains an open challenge across the AI industry.

21. Client Training and Onboarding

Where an AI Service involves Client team members interacting with or managing a new AI system (such as reviewing chatbot conversation logs or approving AI agent actions), DREOS STUDIO will provide reasonable documentation and, where included in the agreed project scope, a walkthrough or training session to help the Client's team understand the system's capabilities, limitations, and recommended operating procedures.

22. Feedback Loops and Iterative Improvement

AI systems often benefit from iterative refinement based on real-world usage patterns. Where a Client wishes to establish an ongoing feedback loop — such as periodically reviewing flagged or low-confidence AI outputs to refine prompts, retrieval sources, or guardrails — this can be scoped as part of an active maintenance agreement. Absent such an arrangement, DREOS STUDIO's responsibility is limited to the AI system's performance as accepted at the conclusion of the applicable Project Agreement's testing period.

23. Glossary of Terms

For clarity, this Section provides plain-language definitions of select terms used throughout these AI Terms:

• "Large Language Model (LLM)" — an AI model trained on large volumes of text data to generate human-like language outputs.
• "Hallucination" — an AI-generated output that is factually incorrect or fabricated, despite appearing plausible or confident.
• "Retrieval-Augmented Generation (RAG)" — a technique that combines a language model with a specific data source (such as a Client's documents) to help ground its responses in relevant, up-to-date information.
• "AI Agent" — an AI-powered system capable of taking multi-step actions toward a goal, potentially including interacting with other software systems.
• "Prompt" — the instructions or input provided to an AI model to guide its output.
• "Foundation Model" — a large, general-purpose AI model (typically developed by a third-party provider) that can be adapted or fine-tuned for specific applications.

24. Environmental and Resource Considerations

The Client acknowledges that training and operating AI models involves computational resources provided by third-party infrastructure and AI providers, and that the environmental footprint of a given AI Service depends largely on the practices of those underlying providers rather than on DREOS STUDIO directly. Clients with specific sustainability reporting requirements related to their technology stack should raise this at the outset of a project so that provider selection can factor in publicly available information about a given provider's infrastructure practices, where relevant.

25. A Note on How We Approach AI Engineering

We build AI systems the same way we build any other software: with attention to edge cases, failure modes, and long-term maintainability, rather than treating AI components as a novelty layered on top of a product. Where a use case calls for deterministic, rule-based logic rather than a probabilistic AI model, we will say so during scoping, since not every problem is best solved with AI. Our goal is to help Clients apply AI where it genuinely adds value, with the guardrails and honest expectations described throughout these AI Terms.

26. Changes to These AI Terms

DREOS STUDIO may update these AI Services Terms from time to time to reflect the evolving nature of AI technology, third-party provider practices, and applicable law. Material changes will be communicated to active Clients with AI Services engagements.

27. Contact

Questions about these AI Services Terms should be directed to: