AI Safety & Acceptable Use Policy

Zero Tolerance for Prompt Injection, Jailbreaking & Guardrail Evasion

SalesChat simulations are strictly dedicated to commercial sales training, objection practice, and skill assessment. Any attempt to manipulate model system prompts, bypass ethical guardrails, induce adversarial behavior, or extract underlying platform heuristics is strictly prohibited and subject to immediate account termination.

1. Purpose & Authorized Use

SalesChat provides dynamic AI-driven buyer personas to enable sales development representatives (SDRs), account executives (AEs), and revenue teams to practice live objection handling and sales negotiation in a safe, constructive environment.

Authorized activities include:

Conducting interactive sales roleplay dialogues across configured product value propositions.
Testing responses against common commercial objections (e.g., budget limits, timing, competitor comparisons).
Reviewing automated turn evaluations, discovery scores, and coaching recommendations generated by our decision engine.
Benchmarking team skill development and rep objection mastery within team workspaces.

2. Prohibited Conduct & Adversarial Prompt Abuse

You agree not to misuse the simulation interface or underlying language models. The following practices are strictly prohibited:

Prompt Injections & Jailbreaks

Crafting inputs intended to override, bypass, or corrupt system instructions, safety boundaries, or persona parameters.

Prompt & Heuristic Extraction

Submitting prompts specifically designed to leak proprietary system prompts, scoring rubrics, or internal evaluation prompts.

Harassment & Abusive Content

Submitting hate speech, slurs, sexually explicit content, discriminatory dialogue, or defamatory material into simulation inputs.

Malicious & Deceptive Generation

Attempting to use the platform to generate malware, phishing templates, fraudulent correspondence, or deceptive outreach campaigns.

Additionally, automated scraping, robotic querying, or bulk extraction of simulation scenarios via non-standard API tools without authorized enterprise access is forbidden.

3. Synthetic Personas & Model Limitations

When using SalesChat, you acknowledge and agree to the following operational characteristics of generative AI:

Synthetic Personas: Buyer personas (e.g., procurement managers, enterprise CTOs, resistant budget owners) are synthetic fictional constructs designed to challenge reps. Any resemblance to real persons, companies, or events is purely coincidental.
No Guarantee of Real-World Outcomes: Winning a simulation or achieving a high score (4 or 5) does not guarantee commercial success in actual sales negotiations.
Hallucination & Variability: Large language models are non-deterministic. While we employ rigorous prompt engineering and temperature constraints, personas may occasionally produce unexpected, counter-factual, or non-optimal responses.
Not Professional Advice: Simulation responses and coaching recommendations do not constitute formal legal, accounting, employment, or financial advice.

4. Technical Guardrails & Enforcement

To maintain service stability, security, and compliance across all customer workspaces, SalesChat enforces programmatic safeguards:

Distributed Rate Limiting: Upstash Redis counters throttle burst requests and protect inference endpoints against automated flooding.

Token & Spend Ceilings: Workspaces operate under predetermined monthly simulation ceilings to prevent runaway inference loops.

Audit Logging & Anomaly Detection: Malicious pattern signatures and repetitive guardrail triggers are flagged for administrative review.

5. Violations & Remediation

If a user or workspace violates this Acceptable Use Policy, SalesChat reserves the right to take remedial action commensurate with the severity of the violation, including:

Issuing formal warnings to workspace administrators.
Temporarily throttling or revoking simulation inference permissions for offending accounts.
Permanently terminating user seats or deleting entire workspaces without refund for egregious, repeated, or malicious exploit attempts.
Cooperating with law enforcement authorities where activities involve illegal conduct.

6. Reporting Violations & Safety Inquiries

If you encounter unexpected model outputs that appear harmful or believe a platform user is violating this policy, please report it promptly to legal@chat2sell.in.

For technical security vulnerabilities, please refer to our Security & Trust page or contact security@chat2sell.in.

Zero Tolerance for Prompt Injection, Jailbreaking & Guardrail Evasion

1. Purpose & Authorized Use

Authorized activities include:

Conducting interactive sales roleplay dialogues across configured product value propositions.

Testing responses against common commercial objections (e.g., budget limits, timing, competitor comparisons).

Reviewing automated turn evaluations, discovery scores, and coaching recommendations generated by our decision engine.

Benchmarking team skill development and rep objection mastery within team workspaces.

2. Prohibited Conduct & Adversarial Prompt Abuse

You agree not to misuse the simulation interface or underlying language models. The following practices are strictly prohibited:

Prompt Injections & Jailbreaks

Crafting inputs intended to override, bypass, or corrupt system instructions, safety boundaries, or persona parameters.

Prompt & Heuristic Extraction

Submitting prompts specifically designed to leak proprietary system prompts, scoring rubrics, or internal evaluation prompts.

Harassment & Abusive Content

Submitting hate speech, slurs, sexually explicit content, discriminatory dialogue, or defamatory material into simulation inputs.

Malicious & Deceptive Generation

Attempting to use the platform to generate malware, phishing templates, fraudulent correspondence, or deceptive outreach campaigns.

Additionally, automated scraping, robotic querying, or bulk extraction of simulation scenarios via non-standard API tools without authorized enterprise access is forbidden.

3. Synthetic Personas & Model Limitations

When using SalesChat, you acknowledge and agree to the following operational characteristics of generative AI:

Synthetic Personas: Buyer personas (e.g., procurement managers, enterprise CTOs, resistant budget owners) are synthetic fictional constructs designed to challenge reps. Any resemblance to real persons, companies, or events is purely coincidental.

No Guarantee of Real-World Outcomes: Winning a simulation or achieving a high score (4 or 5) does not guarantee commercial success in actual sales negotiations.

Hallucination & Variability: Large language models are non-deterministic. While we employ rigorous prompt engineering and temperature constraints, personas may occasionally produce unexpected, counter-factual, or non-optimal responses.

Not Professional Advice: Simulation responses and coaching recommendations do not constitute formal legal, accounting, employment, or financial advice.

4. Technical Guardrails & Enforcement

To maintain service stability, security, and compliance across all customer workspaces, SalesChat enforces programmatic safeguards:

Distributed Rate Limiting: Upstash Redis counters throttle burst requests and protect inference endpoints against automated flooding.

Token & Spend Ceilings: Workspaces operate under predetermined monthly simulation ceilings to prevent runaway inference loops.

Audit Logging & Anomaly Detection: Malicious pattern signatures and repetitive guardrail triggers are flagged for administrative review.

5. Violations & Remediation

If a user or workspace violates this Acceptable Use Policy, SalesChat reserves the right to take remedial action commensurate with the severity of the violation, including:

Issuing formal warnings to workspace administrators.

Temporarily throttling or revoking simulation inference permissions for offending accounts.

Permanently terminating user seats or deleting entire workspaces without refund for egregious, repeated, or malicious exploit attempts.

Cooperating with law enforcement authorities where activities involve illegal conduct.