1. How AI features work

Our AI features use large language models and related machine-learning systems. They generate responses probabilistically from patterns in data — they do not 'look up' guaranteed facts. As a result, outputs can occasionally be incomplete, outdated, or incorrect, even when they sound confident.

2. Not professional advice

AI-generated output is for general information and operational assistance only. It is not legal, financial, medical, tax, or other professional advice, and should not be relied on as a substitute for a qualified professional. Decisions with legal or financial consequences should be reviewed by a person.

3. Human oversight & configuration

AI receptionists and chatbots use the approved knowledge, rules, hours, and hand-off behavior in the configured playbook. Authorized users may review retained and available transcripts and request or make permitted instruction changes; a change applies only after it is validated and confirmed through the applicable control. Transfers, messages, and callbacks remain subject to availability.

CORE is responsible for operating the contracted safeguards, access boundaries, and approval controls with reasonable care. The client is responsible for lawful, accurate business content and instructions and for reviewing output where the workflow assigns that review. Each party remains responsible for its own breach, negligence, or unlawful conduct; no AI system is error-free.

4. Data handling

Conversations and AI-feature data are handled under our Privacy Policy and, where we process your customers' personal data on your behalf, our Data Processing Addendum. CoreTV LLC does not sell this data for monetary consideration. Configured model, speech, transcription, embedding, or retrieval providers process scoped inputs and outputs under CORE's applicable provider and data-protection terms.

5. Usage & billing

AI, voice, transcription, chatbot, messaging, and similar usage may be metered at the managed rate and allowance disclosed before approval, separate from a flat plan fee. Available dashboard usage can lag, synchronize, or be finalized later by a provider; the final invoice and metering evidence control, subject to correction of a verified error. Quoted plan prices exclude tax and variable usage unless expressly included.

6. Limitation of liability

AI output must be reviewed before it is used for a material decision, commitment, publication, or regulated purpose. To the extent permitted by law, CoreTV LLC does not assume responsibility for a user's unreviewed reliance on an output or for an unlawful instruction, but remains responsible for obligations directly assigned to it by the accepted agreement and applicable law. The exclusions and liability limits in the Terms of Service apply; no disclaimer waives a remedy that cannot lawfully be waived.

---

7. Governed AI actions & human decisions

CORE uses a governed action registry, confidence thresholds, role permissions, approval policies, rate limits, audit records, and human escalation to bound AI-assisted actions. Low-risk tasks may run automatically when an administrator has enabled and approved that policy. Material pricing, scope, refund, cancellation, legal acceptance, payment, employment, security, or access decisions require the designated human or participant approval unless a signed agreement and applicable law expressly permit a different process.

A summary, sentiment label, lead score, urgency classification, deal-health or client-success signal, relationship confidence, suggested bundle, draft estimate, or next-action recommendation is an aid to a human workflow. It can be wrong and should not be treated as an objective fact about a person.

When the separately governed voluntary applicant questionnaire is enabled, its race or ethnicity, gender, veteran, and disability responses are not supplied to AI systems, conversation scores, ranking, interviewing, or ordinary application review. CORE does not use an AI output or conversation score as the sole basis for a final employment or engagement decision.

A learning or capability graph may recommend content or infer a knowledge gap from role requirements, submitted evidence, practice results, and authorized work records. It can be incomplete or wrong and does not by itself award or revoke a credential, assign work, set compensation, or make an employment decision.

8. AI receptionist, transcription & emergencies

AI receptionists and sales or support chat can collect a request, answer from approved knowledge, present available booking times, prepare a draft estimate, create a ticket, transfer a call, or schedule a callback. The system may generate live captions or transcripts for the caller and staff after the legally required notice and consent.

AI voice and chat are not emergency, medical, legal, crisis, or public-safety services. They cannot guarantee that a person will answer a transfer or callback. A caller who needs immediate emergency help must contact the appropriate emergency service directly.

9. Estimate, negotiation & proposal limits

AI may explain catalog prices, identify missing scope, compare tiers, suggest tradeoffs, or prepare a proposal and review note. It may not create an undisclosed discount, change a contracted price, guarantee a launch date or result, or bind CORE outside the formal quote and approval process. Budget or ROI illustrations depend on assumptions and are not financial forecasts.

10. AI data sources, subprocessors & training

AI output may use the information supplied in the current conversation, files or URLs submitted for the requested intake, approved CORE catalog and knowledge content, linked account or project records the user is authorized to access, operational and relationship signals, and third-party model, extraction, retrieval, or transcription services. Users must not submit data they lack authority to share.

An AI-generated extraction, citation, summary, architecture node, timeline, cost explanation, or source reference may omit, misread, or misattribute information. The user and assigned reviewer must verify material facts against the original source and approved commercial record. Passwords, authentication codes, private keys, and payment-card data must use the designated protected workflow rather than an AI prompt or ordinary upload.

CORE does not use private client or applicant content to train a public foundation model. Model and speech providers process inputs and outputs to supply the requested feature under the applicable service and data-protection terms. Any client-specific model training or materially different secondary use requires a separate written scope and any consent required by law.

11. AI security, prompt attacks & misuse

CORE filters and bounds tool access, but no defense eliminates every prompt injection, malicious file, poisoned source, or model error. System instructions, credentials, hidden policies, private context, and other users' data are confidential. Attempts to extract, override, or expose them are prohibited and may be blocked, logged, investigated, and reported.

12. Automated security-signal limits

Public-request protections may deterministically compare an IP address, network, or email domain with configured deny, allow, and disposable-domain rules. A rule match can reject or rate-limit the request but is not an AI conclusion about identity, character, creditworthiness, employment, or legal fault. CORE does not use that match by itself to make a binding price, hiring, eligibility, or other high-impact decision.

AI may summarize an incident or recommend investigation only within the governed action and human-review controls described above. It does not receive permission to expose the underlying denylist, reveal another user's request, override an allowed-domain exception, or silently convert a security signal into a permanent account sanction.