KokuaOS — Responsible AI Overview

A plain-English summary of how KokuaOS approaches responsible, transparent, and human-accountable use of AI.

Capitalized terms have the meanings in the KokuaOS Definitions Schedule (kokuaos.com/legal/definitions).

1. Our commitment

KokuaOS builds AI Employees that talk with real people across voice, messaging, and other Communication Channels. We take seriously the responsibility that comes with that. Our goal is AI that is useful, transparent, kept under human oversight, and used within appropriate limits. This page summarizes our approach in plain English. For the binding commitments and definitions, see the Responsible AI Policy (kokuaos.com/legal/responsible-ai) and the AI Services & AI Consumption Policy (kokuaos.com/legal/ai-consumption), which control if anything here appears to differ.

Responsible AI on the Platform is a shared responsibility: KokuaOS provides the capabilities and guardrails, and Customers decide how to configure and deploy AI Employees for their own use cases.

2. Human oversight and accountability

AI Employees are tools that assist people; they are not a replacement for human judgment. The Platform is designed to keep a human in control — Customers configure what an AI Employee may and may not do, set escalation and hand-off to human staff, and retain the ability to review, correct, and override AI behavior. Customers must maintain qualified human review before relying on AI Output for any legal, medical, financial, employment, safety, eligibility, or other high-risk or legally significant decision.

3. Accuracy limitations and honesty about hallucinations

AI Output can be inaccurate, incomplete, outdated, or biased, and generative systems can "hallucinate" — produce content that sounds authoritative but is wrong. We are candid about this because it matters for how AI Employees should be deployed. AI Output is not professional, legal, medical, financial, or other regulated advice. We design features to reduce errors — such as grounding responses in a Customer's Knowledge Base and enabling escalation — but no configuration eliminates the risk, and Customers are responsible for reviewing AI Output before relying on it.

4. Transparency and AI disclosure

We support telling people when they are interacting with an AI system. The Compliance Shield lets Customers configure AI-identity disclosures, recording notices, and consent prompts appropriate to their channel and jurisdiction. Customers are responsible for determining and enabling the disclosures that their laws and policies require.

5. Fairness and bias awareness

AI systems can reflect biases present in data and models. We are mindful of fairness in how we build the Platform, and we encourage Customers to consider fairness in how they deploy AI Employees — particularly where outcomes affect access to employment, credit, housing, healthcare, or other significant opportunities. Customers deploying AI in these contexts are responsible for appropriate testing, human review, and compliance with anti-discrimination and consumer-protection law.

6. Access controls and security

Responsible AI depends on strong security. Access to AI Employees, configurations, prompts, and data is governed by role-based access control and the safeguards described in our Security Overview (kokuaos.com/trust/security), including encryption in transit and at rest, tenant isolation, and field-level protection of secrets.

7. Testing, evaluation, and monitoring

We test and monitor the Platform for quality, safety, and abuse, and we provide Customers with a Sandbox for configuring and evaluating AI Employees before Production. (Sandbox is not a free-usage environment — testing activity may generate billable AI Consumption.) We encourage Customers to test their AI Employees against realistic scenarios, review transcripts, and monitor performance on an ongoing basis.

8. High-risk use restrictions

Some uses are inappropriate for automated AI handling or require heightened safeguards. Customers must not deploy AI Employees to make final, unreviewed high-risk decisions — including in legal, medical, financial, employment, safety, and eligibility contexts — without qualified human review. Certain uses are prohibited outright under the Acceptable Use Policy (kokuaos.com/legal/aup), and regulated data such as PHI and full cardholder data is restricted as described in the Compliance Overview (kokuaos.com/trust/compliance).

9. Data and privacy in AI

• No training on Customer Content without authorization. KokuaOS does not train foundation models on Customer Content without the Customer's explicit written authorization.
• Aggregated and de-identified analytics. KokuaOS may use aggregated and de-identified data and analytics that cannot reasonably identify a person or Customer, to operate and improve the Services.
• BYO and managed providers. AI processing runs through KokuaOS-managed model providers or, where a Service Provider brings its own, its BYO Providers. Provider handling of data is addressed in the Privacy Policy, the DPA, and the BYO Provider Policy.

10. Your responsibilities

When deploying AI Employees, Customers are responsible for, among other things:

• Configuring AI Employees, prompts, and Knowledge Bases lawfully and appropriately for the use case.
• Enabling required AI, recording, and consent disclosures.
• Maintaining human review for high-risk or legally significant decisions.
• Avoiding prohibited and restricted uses, and restricted data types (PHI, full card data).
• Testing, monitoring, and correcting AI Employee behavior over time.
• Training their workforce on appropriate use.

11. Reporting concerns

If you encounter AI Output or behavior that appears harmful, unsafe, biased, or non-compliant, please report it so we can investigate. Customers and their end users can raise concerns through their KokuaOS administrator or by contacting trust@kokuaos.com. Suspected security issues should go to security@kokuaos.com under the Vulnerability Disclosure Policy (kokuaos.com/legal/vulnerability-disclosure).

12. No guarantee of error-free output

Artificial intelligence is probabilistic and imperfect. KokuaOS does not and cannot guarantee that AI Output will be accurate, complete, current, unbiased, or error-free, and disclaims liability for reliance on AI Output to the fullest extent permitted by law. This overview describes our approach as of the date above, is provided for information only, and may change as the technology and our program evolve. It does not modify any agreement and does not create any warranty or commitment; the binding terms are the KokuaOS Agreement and the Responsible AI Policy.