Terms of service

1. Overview

These Terms of Service govern your use of PromptSafe^®, operated by Capital Consult Ltd (trading as Sacher AI), a company registered in England and Wales, company number 06800173, registered office 6/28a Wincott Street, London, SE11 4NT. By creating an account or using the platform, you agree to these terms in full.

If you have any questions, contact us at info@sacher.ai.

2. Eligibility

PromptSafe is intended solely for business and professional use by users aged 18 or older.

3. Accounts

• You must provide accurate information when registering and keep your account credentials secure.
• Account access is personal and non-transferable. You are responsible for all activity under your account.
• You must notify us immediately at info@sacher.ai if you believe your account has been compromised.
• We reserve the right to suspend or terminate accounts for breach of these terms, abuse of the platform, or non-payment.

4. Free trial

• New users receive a 7 day free trial with full platform access and 30 PS tokens per day. No credit card is required to start a trial.
• At the end of the trial period, access to the platform will be suspended unless you purchase a Pack.
• Data created during a free trial (agents, personas, evaluators, simulations) is retained for a limited period after the trial ends so you can resume access. It is deleted if no Pack is purchased.
• You may not create multiple accounts to circumvent trial limits or rate limits. Doing so may result in immediate account termination.
• We reserve the right to modify, limit, or discontinue the free trial offer at any time without notice.

5. Packs, PS tokens, and payment

After the free trial, paid usage is provided through Packs. One Pack is £99 and includes 2,000 PS tokens, valid for 90 days from purchase. Pricing is published on the website and may change with reasonable notice.

Rolling expiry across multiple Packs. PS tokens from every Pack you hold are pooled into a single workspace balance. The expiry on the entire balance resets to 90 days from your most recent Pack purchase, no matter when earlier Packs were purchased. Buying a new Pack extends the expiry on the tokens you already have. Unused PS tokens expire automatically in accordance with the applicable Pack terms.

Payments are processed by Stripe. By purchasing a Pack you also agree to Stripe's terms of service.

PS tokens are a contractual usage metric solely for accessing PromptSafe services and do not constitute stored monetary value, electronic money, digital currency, or transferable property. The PS token is the only user facing usage unit. We reserve the right to change the underlying models and infrastructure that power the service.

Except where required by applicable law, Packs, PS tokens, and associated fees are non-refundable, non-transferable, and have no cash redemption value.

We may introduce subscription plans, recurring billing models, or alternative pricing structures in the future. Any such changes will apply prospectively and will not affect previously purchased Packs except where expressly stated at the time of purchase.

6. Acceptable use

You agree not to use PromptSafe to:

• Test, build, or deploy AI agents designed to cause harm, facilitate illegal activity, or exploit vulnerable individuals.
• Process real personal data of identifiable individuals without an appropriate legal basis.
• Reverse-engineer, scrape, copy, or reproduce any part of the platform or its underlying systems.
• Circumvent rate limits, usage caps, or access controls through any means, including by creating multiple accounts or workspaces.
• Use the platform in any way that violates applicable laws or regulations.
• Use the platform in any way that could damage, disable, overload, interfere with, or impair the platform or other users' access.

We reserve the right to suspend or terminate access immediately if we become aware of a violation of these provisions.

7. Prohibited uploads and content

Customers must not upload or process content through PromptSafe unless they have all necessary rights, permissions, notices, and lawful bases required under applicable law.

Customers must not upload:

• Unlawful content.
• Regulated health records without authorisation.
• Personal data without lawful basis.
• Confidential third-party information unlawfully.
• Content infringing intellectual property or privacy rights.

8. Customer responsibility

Customers remain solely responsible for:

• Their AI systems.
• Deployment decisions.
• Monitoring decisions.
• Governance processes.
• Regulatory assessments.
• Clinical review obligations.
• Legal compliance.
• Outputs generated by their systems.

9. Your workspace content

You retain ownership of the agents, personas, evaluators, and other content you create within the platform. By using the platform, you grant us a limited licence to process your content to provide the service and, as set out in section 10, to develop, validate, improve, and benchmark our evaluation technology using Evaluation Assets. You are responsible for ensuring you have the appropriate rights to any content you upload or input into the platform.

10. Workspace isolation and model training

Customer workspace content is logically segregated from other customer environments and is not intentionally shared across customer workspaces.

We distinguish between Customer Product Assets and Evaluation Assets. Customer Product Assets means materials that describe, configure or reveal the Customer's own AI system or proprietary information, including agent prompts, system prompts, agent instructions, agent documentation, uploaded knowledge base or RAG material, proprietary workflows, business logic and configuration. Evaluation Assets means materials created or used in the PromptSafe evaluation process, including personas (PromptSafe-generated and customer-created), evaluators (PromptSafe-generated and customer-created), synthetic simulation conversations, customer-entered role-play content, evaluator scores, outputs, reasoning and labels, quality scores, evidence metrics, benchmark results, derived metadata and aggregate statistics.

We do not use Customer Product Assets or workspace content to train third party or general purpose foundation AI models. We do not use Customer Product Assets for research or product improvement by default; any exceptional use requires a clear legal basis, internal governance approval, and appropriate transparency.

Subject to applicable law and the safeguards described in our Privacy policy, we may use Evaluation Assets, derived metadata, aggregated metrics, and appropriately handled customer-derived information to develop, validate, improve, and benchmark our proprietary evaluation methodologies, algorithms, classifiers, persona generation systems, benchmarking technologies, and related artificial intelligence systems. Such processing will not be used to identify individual customers or users, and will not be used to train third party or general purpose foundation AI models.

When we process your workspace content to deliver the service, we act as a processor on your instructions. When we use Evaluation Assets for our own research and product improvement, we act as a controller for that processing, as set out in our Privacy policy.

We do not routinely access customer workspace content. Limited access may occur where reasonably necessary for customer support, security investigation, legal compliance, platform maintenance, or prevention of misuse.

On the self-serve tier, Sacher AI may add curated example content (personas, evaluators) to your workspace where it is sector-relevant. Added items are labelled as examples so they are distinguishable from your own work. Sacher AI does not intentionally modify or delete content you have created except where required for platform operation, legal compliance, security, abuse prevention, or at your request.

For enterprise engagements, Sacher AI access to the workspace forms part of the agreed engagement and is documented in the order form or statement of work.

11. Data processing

Where required by applicable data protection law, the parties may enter into a separate data processing agreement governing processing of customer workspace content. Our Privacy policy sets out further detail on our role as controller and processor and on the service providers we work with.

12. Third-party AI providers

Important. PromptSafe integrates with third-party AI providers including OpenAI, Anthropic, and xAI. If you use Sacher AI-provided model access, conversation data is transmitted to these providers for processing. If you connect your own API key on the enterprise tier, your use of those providers is governed by your own agreement with them.

Sacher AI is not responsible for the outputs, accuracy, availability, or behaviour of third-party AI models. Sacher AI does not control and cannot guarantee the ongoing availability, behaviour, or outputs of third-party AI models. Third-party providers may change their models, pricing, or terms of service at any time. We will make reasonable efforts to communicate material changes to users.

13. Rate limits and fair use

The platform operates under daily PS token limits on the free trial, and Pack-based usage on the paid tier. These limits exist to ensure fair access for all users and to prevent abuse. If you exceed the applicable limit, access may be suspended until the following day or until a new Pack is purchased. Limits may be adjusted as the platform develops and will be communicated to users.

14. Managed service

Where Sacher AI provides a managed service (including running evaluations on your behalf, building bespoke personas, or acting as an independent AI quality partner), this is provided under a separate services agreement. These Terms of Service cover platform access only and do not govern any managed service engagement.

15. Sensitive content within the platform

PromptSafe contains synthetic personas, conversation transcripts, evaluators, and example agents that are designed to test how conversational AI behaves under realistic and adversarial pressure. To do that job, the platform necessarily contains references to and simulations of sensitive subject matter. This may include, but is not limited to: clinical risk and decline; mental health distress; substance use and addiction; coercive or abusive dynamics; discriminatory or harmful language; and other themes a user may find uncomfortable or distressing.

All such content is synthetic and exists solely to support safety testing. It does not reflect the views of Capital Consult Ltd, Sacher AI, or any partner organisation.

By creating an account and using PromptSafe, you confirm that you understand the nature of the content you may encounter, that you are accessing the platform in a professional testing capacity, and that you may stop using the platform at any time. You agree not to share, redistribute, or display PromptSafe content outside the legitimate testing purposes for which it was created.

If you become aware of content that you believe is genuinely harmful, defamatory, or outside the platform's stated purpose, please notify us at info@sacher.ai and we will review it promptly.

16. Disclaimers

16.1 What PromptSafe is, and is not

PromptSafe provides evaluative tooling intended to support internal testing, governance, and quality assurance workflows.

PromptSafe does not certify, guarantee, warrant, or independently verify the safety, effectiveness, legality, regulatory compliance, or clinical suitability of any AI system. Use of PromptSafe does not by itself establish compliance with any legal, regulatory, certification, assurance, audit, or governance framework.

PromptSafe is not intended to diagnose, treat, cure, mitigate, monitor, or prevent disease or medical conditions. PromptSafe is not a medical device, clinical decision support system, or regulatory certification service.

16.2 Evaluation scores are not safety certification

This section is important and should be read carefully before using the platform.

A score of 80%, 100%, or any percentage produced by a PromptSafe evaluation does not mean your AI agent is safe, compliant, fit for deployment, or free from harmful behaviours. Evaluation scores are probabilistic outputs generated by LLM-based evaluator agents. They reflect how your agent performed across a defined set of simulated conversations at a specific point in time, against the criteria you or we have defined.

Evaluation scores should be used as a relative and directional indicator, not as an absolute measure of safety or quality. Specifically:

• A high score does not guarantee your agent will behave safely or correctly with real users in production.
• A score of 100% on a PromptSafe evaluation does not mean your agent is 100% safe. It means it performed well against the evaluated criteria in the simulated conversations run at that time.
• Scores are most meaningful when tracked over time. If you update your agent and the score improves, that is evidence of relative improvement against your defined criteria. If the score falls, that is a signal to investigate.
• The value of evaluations lies primarily in identifying where and why your agent is underperforming, so you can make targeted improvements.
• Simulated conversations do not replicate the full range of real-world interactions. Real users will behave in ways that are not covered by any set of simulated personas.

16.3 International regulatory frameworks

PromptSafe testing records may form part of supporting documentation within AI governance programmes aligned to international frameworks including ISO 42001, the IMDRF Good Machine Learning Practice (GMLP) guiding principles, the EU AI Act, and the AI guidance frameworks of the FDA (US), Health Canada, the MHRA (UK), and the TGA (Australia). In all cases, PromptSafe alone does not satisfy the requirements of these frameworks, which encompass statistical testing standards, clinical validation, data representativeness, and broader quality management system requirements that go beyond what PromptSafe provides. You should seek independent regulatory and legal advice appropriate to your jurisdiction and the nature of your AI product.

16.4 Platform availability and evolution

The platform is provided “as is”. We make no warranty that the platform will be error-free, uninterrupted, or suitable for any particular purpose. Outputs of simulations are indicative only and should inform, not replace, human review and professional judgement.

PromptSafe may evolve over time and certain features may be modified, suspended, replaced, or discontinued.

17. Limitation of liability

Nothing in these Terms excludes or limits liability for:

• Death or personal injury caused by negligence.
• Fraud or fraudulent misrepresentation.
• Liability which cannot legally be excluded.

Subject to the above, and to the fullest extent permitted by law, Sacher AI's total liability to you for any claim arising from or related to these terms or your use of the platform shall not exceed the total fees you have paid to us in the 12 months preceding the claim.

To the maximum extent permitted by law, we exclude liability for indirect, incidental, consequential, special, punitive, or exemplary damages, including loss of profits, revenue, goodwill, business opportunity, or anticipated savings.

Sacher AI shall have no liability arising from deployment, clinical, operational, governance, or commercial decisions made by customers based on PromptSafe outputs.

18. Force majeure

Sacher AI shall not be liable for delay or failure to perform obligations caused by events beyond its reasonable control.

19. Assignment

You may not assign or transfer your rights under these Terms without our prior written consent.

20. Intellectual property

• PromptSafe^® is a registered trademark of Capital Consult Ltd.
• You retain ownership of your Customer Product Assets (your agent prompts, system instructions, uploaded documentation, knowledge bases, RAG content, proprietary workflows, and configuration), and you retain any intellectual property rights that subsist in the Evaluation Assets you supply.
• "Derived Evaluation Intelligence" means the analyses, classifications, benchmarks, taxonomies, statistical models, evaluation methodologies, derived metadata, and other research outputs that Sacher AI creates through operating the platform. It does not include your Customer Product Assets or the Evaluation Assets you supply.
• Sacher AI owns all intellectual property rights in PromptSafe, including the platform code and design, its evaluation methodologies, classifiers, benchmarking systems, persona and evaluator libraries, derived metadata, aggregated statistics, taxonomies, research outputs, and other Derived Evaluation Intelligence created through operation of the platform.
• You are granted a limited, non-exclusive, non-transferable licence to use the platform for your own internal AI development and testing purposes only.
• Nothing in these terms transfers any intellectual property rights to you.

21. Confidentiality

Each party will treat confidential information disclosed by the other party as confidential and will not disclose it to third parties without the other party's consent, except as required to provide the service or by law.

22. Changes to these terms

We may update these Terms of Service from time to time. We will provide reasonable notice of material changes by email or via a notice on the platform. Continued use of the platform after the effective date of any changes constitutes acceptance of the updated terms.

23. Governing law

These terms are governed by the laws of England and Wales. Any disputes arising from these terms or your use of the platform shall be subject to the exclusive jurisdiction of the English courts.

For enterprise customers, the governing law, jurisdiction, and dispute resolution provisions may be varied by a separate signed agreement. Where an enterprise order form or master agreement specifies different terms, those terms prevail over this section for that customer.

24. Contact

Capital Consult Ltd (trading as Sacher AI)
Email: info@sacher.ai
Website: sacher.ai

Related: Privacy policy, Cookie policy.