What Masheev does not guarantee
AI output is probabilistic. Masheev does not guarantee that a generated answer is accurate, complete, current, suitable for a particular purpose, compliant in the customer's jurisdiction, or accepted by a provider. Customers remain responsible for their instructions, knowledge, channels, recipients, notices, consent, and use of AI output.
Guardrails and escalation reduce risk, but they do not remove the need for human judgement. A customer must test the real use case, monitor production outcomes, correct unsafe behavior, and stop an affected path when continued automation may cause harm.
The safe operating loop
Each control depends on the one before it. Disclosure without constraints is not enough, and escalation without a staffed owner is not a working safeguard.
Disclose
Tell people when AI is involved
Give end users clear notice when they are interacting with AI where law, channel rules, context, or reasonable expectations require it. Never present an AI agent, synthetic voice, or AI-written message as a real person.
Constrain
Limit the agent to approved work
Provide current business knowledge, narrow instructions, permitted tools, spending limits, and explicit escalation rules. Do not give an agent authority that the person configuring it does not have.
Observe
Review outcomes and failure signals
Monitor conversations, tool results, provider delivery, escalations, audit events, and billing. A sent or generated response is not proof that it was correct, delivered, lawful, or appropriate.
Escalate
Move uncertain work to a person
Masheev can create escalations for low confidence, configured keywords, tool failure, repeated failure, or an explicit request for a person. Customers must assign and staff the people who receive them.
Approve
Let a person approve what the agent learns
A resolved escalation can produce a proposed question and answer. It enters the knowledge base only after a person reviews and approves it. Do not approve sensitive, temporary, unverified, or customer-specific facts as reusable knowledge.
When a person must take over
A person asks for one
Honor an explicit request for human assistance. Do not trap someone in an automated loop.
The answer is uncertain or outside approved knowledge
Escalate instead of inventing a policy, price, availability, diagnosis, commitment, or provider state.
A decision could significantly affect a person
Do not use Masheev for solely automated legal or similarly significant decisions without appropriate human review and intervention.
Professional judgement is required
Agents do not diagnose or give medical, legal, or financial advice; they share the business's own published information and suggest a qualified professional. Anyone describing a medical emergency or immediate danger is told to call their local emergency number.
Identity, payment, privacy, or security is in question
A person must verify authority through an approved process. The agent and support team must never request passwords, one-time codes, API keys, or full card details.
Human handoff is not an emergency service and does not guarantee an immediate live response. Each Masheev customer is responsible for assigning eligible team members, setting coverage, reviewing open escalations, and giving end users another suitable route when timing or safety requires one.
Uses Masheev does not permit
The full Acceptable Use Policy controls. In summary, do not use Masheev to create, facilitate, or scale:
- Fraud, phishing, credential harvesting, impersonation, scams, or deceptive synthetic media
- Harassment, discrimination, exploitation, unlawful surveillance, or illegal content and activity
- Unsolicited bulk messaging, caller-ID manipulation, opt-out evasion, or communications without required consent
- Bypassing safety controls, extracting system prompts or model data, prompt injection against other people, or unauthorized access
- High-risk or prohibited automated decisions, professional advice, therapy or companion use, or regulated data use without the required written agreement
- Protected health information under HIPAA: Masheev does not sign Business Associate Agreements, so US practices covered by HIPAA must not use it with patient information
Masheev may investigate, limit, suspend, or terminate use that violates the policy or creates unacceptable risk. Report suspected abuse to abuse@masheev.com. Do not include passwords, tokens, one-time codes, full payment details, or unnecessary personal data.
If you are speaking with a Masheev-powered agent
- Ask for a person when you do not want to continue with automated assistance.
- Do not send passwords, security codes, full card details, or identity documents.
- Verify important prices, policies, appointments, and commitments with the business.
- Contact the business directly for urgent, professional, safety-critical, account, privacy, or payment matters.
For questions about Masheev's own service, visit the Help Centre. For privacy rights involving Masheev, use the privacy request page.