Klinko AI Feature Guidelines

1. Understand AI limitations

Review important sources, contradictory evidence, sample limitations, dates, and assumptions before relying on a conclusion.

• Output may contain factual errors, missing context, outdated information, fabricated details, or bias.
• A confidence label or ranking is a decision aid, not a statistical guarantee.
• Patterns in market evidence may not represent an entire population or prove causation, prevalence, or willingness to pay.

2. Use lawful, appropriate inputs

Submit only information you have the right and lawful basis to process. Remove unnecessary personal, confidential, financial, health, biometric, credential, or other sensitive information before uploading material or writing prompts.

3. Do not invent evidence

Never represent generated examples, synthetic personas, inferred statements, or model-created quotes as interviews, observed behavior, verified customer testimony, or measured facts. Keep observation, inference, recommendation, and hypothetical illustration clearly separated.

4. Prohibited uses

• Unlawful discrimination, social scoring, or decisions based on sensitive or protected characteristics.
• Re-identification, doxxing, stalking, invasive profiling, or surveillance of individuals.
• Fraud, impersonation, phishing, manipulation, malicious persuasion, or deceptive targeting.
• Circumventing safeguards, access controls, rate limits, source restrictions, or legally required disclosures.
• Generating instructions or analysis intended to facilitate serious wrongdoing or real-world harm.

5. High-impact decisions

Do not use Klinko output as the sole basis for decisions about employment, credit, insurance, housing, education, healthcare, law enforcement, immigration, or another area that significantly affects a person’s rights or opportunities. Such uses require a lawful basis, qualified review, meaningful human oversight, and an appropriate appeal process.

6. Customer and market research ethics

Use proportional methods, respect source context and community expectations, and avoid extracting more information than needed. Do not treat louder, more active, or more visible groups as automatically representative. Validate important hypotheses with direct research, experiments, first-party behavior, or other suitable evidence.

7. Publishing and disclosure

When sharing consequential findings, identify material sources, time range, scope, limitations, and the role of AI. Do not hide contradictory evidence or imply precision that the underlying information does not support. Correct important errors promptly.

8. Intellectual property and source rights

Check whether input and output are subject to copyright, database rights, trademarks, confidentiality, platform terms, or other restrictions before reproducing or commercializing them. Citations do not by themselves grant permission.

9. Signal Watch and automated monitoring

Signal Watch compares new market evidence with a saved decision baseline and may produce alerts or rerank an audience. It can miss changes, overstate weak patterns, or be delayed by evidence and delivery conditions. Treat each alert as a prompt to inspect the rationale and decide whether further validation is needed—not as an automatic instruction to change strategy.

10. Safeguards and reporting

We may block requests, limit features, flag output, require additional review, or restrict accounts when we identify safety, legal, or integrity risks. If output is harmful, privacy-invasive, infringing, or clearly incorrect, stop relying on or distributing it and report it through product feedback or business@klinko.ai.