What the assistant is allowed to do, what it will not do, and who owns what it writes. The governing idea is the Assistant-Operator doctrine: the AI drafts and interprets, the organizer remains the operator of record and the final authority.
This Responsible AI & Usage Policy (“AI Policy”) governs the development, deployment, and use of artificial intelligence, machine learning, and automated processing technologies (collectively, “AI Systems”) within the MustardSMS platform, a brand owned and operated by Hexoglyph Holdings, LLC (“Hexoglyph”).
At MustardSMS, we are committed to the principle that AI should serve as a force multiplier for human coordination, never a replacement for human accountability.
1
The Assistant-Operator doctrine
The Service is architected as a Human-Directed Algorithmic Support System, categorized internally as our “Assistant-Operator Model.”
Advisory logic: Our AI Systems function strictly as administrative assistants. They are designed to interpret intent, summarize data, and suggest actions (the “Assistant”).
Final authority: The User (the “Operator”) retains exclusive, non-delegable authority over all system outputs. No automated interpretation becomes a “persistent record” or “outbound communication” without the explicit affirmative action or approval of the Operator.
Human-in-the-loop (HITL): We mandate HITL protocols for any action that impacts the rights, obligations, or logistical commitments of End Recipients.
2
Transparency and multi-state compliance
In accordance with emerging U.S. state regulations, we provide the following disclosures:
AI interaction notice: Users and, where applicable, End Recipients are notified that the Service utilizes AI to process and interpret communications.
Algorithmic disclosure: Our AI Systems use probabilistic modeling to interpret natural language. These models are not “truth engines” but are “intent-prediction tools.”
No statutory profiling: The Service does not engage in “Profiling” in furtherance of decisions that produce legal or similarly significant effects concerning a consumer.
3
Data privacy and zero-retention governance
No model training: Hexoglyph does not use User Data, Message Content, or proprietary event lists to train, retrain, or improve public or foundational third-party AI models.
Enterprise-tier API guardrails: We utilize enterprise-level AI interfaces that are contractually bound by Zero-Retention policies.
Data minimization: We utilize “context-stripping” techniques to ensure that only the minimum necessary data required for logistical interpretation is processed.
4
Ownership of AI-generated output
User ownership: as between the User and Hexoglyph, the User owns all “Output” generated by the AI Systems based on the User's “Input” (Message Content).
License back to Company: the User grants Hexoglyph a limited, non-exclusive license to process such Output solely to provide the Service.
5
Multimodal processing: voice and image analysis
Voice analysis: If a User utilizes voice features, AI may be used for speech-to-text transcription and sentiment analysis. These outputs are advisory.
FCC/TCPA compliance (voice): Users are strictly prohibited from using the Service to initiate outbound calls using AI-generated or “cloned” synthetic voices without heightened consent.
Image analysis (OCR): AI may be used to extract text from images (e.g., reading an RSVP card).
Biometric prohibition: MustardSMS does not use AI for biometric identification. We do not utilize facial recognition or fingerprint analysis.
6
Telecommunications safety and SHAFT compliance
Automated content filtering: Our AI Systems are configured to detect and flag violations of carrier “SHAFT” (Sex, Hate, Alcohol, Firearms, Tobacco) policies.
Abuse prevention: AI logic is used to identify and halt “burst” messaging or patterns indicative of spam or harassment.
Neutral conduit: Hexoglyph remains a neutral conduit; AI safeguards do not shift liability for Message Content from the Operator.
7
Reliability, hallucination and risk assumption
Generative AI is subject to “hallucinations,” the generation of factually incorrect or illogical information.
Mandatory verification: Operators are contractually and ethically required to verify the accuracy of AI-suggested summaries before relying on them.
Assumption of risk: Users acknowledge that AI interpretation is inherently probabilistic and assume all risks associated with AI inaccuracies not caught during review.
8
Prohibited AI behaviors
Users may not utilize or attempt to utilize the MustardSMS AI Systems to:
- Generate “Deepfake” or deceptive audio/text content.
- Automate harassment or high-frequency “bot-like” interactions.
- Reverse-engineer or “jailbreak” the underlying AI logic.
9
Policy updates and regulatory cooperation
As federal and state AI regulations evolve, Hexoglyph will update this policy accordingly. We reserve the right to modify or restrict AI features to ensure ongoing compliance.
10
Contact information
Privacy & AI Policy Officer, Hexoglyph Holdings, LLC.
Submit inquiries via our Contact page
