Terms of Service

tomphillips.uk is a portfolio website with interactive demos. SentinelProxy is a software-as-a-service platform that ingests logs from your AI agents, evaluates each exchange against configurable guardrails (such as prompt injection, PII leakage, hallucination and competitor mentions), and proposes hardened system prompts. It runs asynchronously — off your application's critical path. The Service is provided on an “as is” basis (see §10 and §11) and is currently offered in beta.

• You must be 18 or older to create an account.
• You must provide accurate registration details and keep them up to date.
• You're responsible for keeping your password and API keys secret, and for all activity that occurs under your account or keys.
• Tell us promptly at tom@tomphillips.uk if you suspect unauthorised access.
• One account per person or organisation unless otherwise agreed.

You retain all rights to the data you send to the Service — system prompts, user inputs, model responses and related metadata (“Customer Data”). You grant us a limited licence to process Customer Data solely to provide and improve the Service for you.

For Customer Data, we act as your data processor and you are the controller; our processing is governed by these terms and the Privacy Policy, which together form our data-processing terms. You are responsible for having the right to send Customer Data to us and for not sending data you are not permitted to share.

• We do not sell Customer Data, and we do not use it to train machine-learning models.
• We process Customer Data only on your documented instructions (i.e. to run the evaluations and tools you trigger).
• SentinelProxy's own software, detectors, prompts and generated fixes remain our intellectual property.

• Don't send unlawful content, or any data you don't have the right to process.
• Don't attempt to interfere with, attack, overload, or probe the Service for vulnerabilities. Responsible disclosure to tom@tomphillips.uk is welcome and won't lead to legal action against good-faith researchers.
• Don't exceed published rate limits or attempt to circumvent them.
• Don't use the Service to build a competing product, or resell it without agreement.
• Don't impersonate others or misrepresent your identity.
• We may suspend or terminate accounts that materially breach these terms.

You may optionally supply your own Anthropic API key so that the Claude calls SentinelProxy makes on your behalf bill to your own Anthropic account. If you do:

• You are responsible for your Anthropic account, its costs, usage limits, and Anthropic's own terms.
• We store your key encrypted at rest and use it only to make model calls for your workspace; we never display it again or expose it to your browser.
• We are not responsible for charges on, or the availability or behaviour of, your Anthropic account or the underlying models.
• You can remove your key at any time, which disables BYOK for your workspace.

The Service offers a free tier and paid subscription tiers with higher limits. Where you take a paid plan:

• Prices are shown in GBP; applicable taxes are added at checkout.
• Payments are processed by Stripe — card data never touches our servers.
• Subscriptions renew until cancelled; you can cancel at any time, effective at the end of the current billing period.
• We may change pricing or plan limits with reasonable notice; changes don't apply retroactively to a period you've already paid for.

Where you've paid for a subscription, refunds are handled per our Refund Policy — or simply email tom@tomphillips.uk. Your statutory rights are unaffected.

We aim for high availability but don't guarantee uninterrupted access; brief planned downtime may occur for maintenance. As a beta product, features may change, and limits, detectors and models may be adjusted over time. Because SentinelProxy is asynchronous, an interruption to it does not block your own application.

You may delete your workspace at any time from Account → Danger zone, which permanently deletes all your SentinelProxy data (projects, API keys, logs, violations and fixes). We may suspend or terminate access for material breach of these terms, fraud, or abuse; where reasonable and proportionate we'll try to contact you first.

SentinelProxy is a tool that helps you monitor and improve the safety and reliability of your AI agents. It does not detect every possible issue, and it is not a guarantee of compliance, security, or correct model behaviour. You remain responsible for your agents' outputs and for meeting your own legal and regulatory obligations. Evaluations and suggested fixes are generated by automated systems and should be reviewed before you rely on them.

To the maximum extent permitted by law, the Service is provided “as is” without warranties of any kind. We exclude liability for indirect, consequential, or incidental losses, and for losses arising from third-party model providers.

Our total aggregate liability for any claim arising from these terms or the Service is limited to the greater of the amount you paid us in the 12 months preceding the claim, or £100.

Nothing in these terms excludes liability that cannot be excluded under UK law, including for death or personal injury caused by negligence, fraud, or statutory rights that cannot be waived.

We may update these terms from time to time. Material changes will be flagged in-product or notified by email if you have an account. Continued use after a change takes effect constitutes acceptance.

These terms are governed by the laws of England and Wales, and disputes will be heard by its courts (except that consumers resident elsewhere retain any local mandatory rights). For questions, email tom@tomphillips.uk; for privacy specifically, see the Privacy Policy.