PART 01 / PRIVACY NOTICE
The information we use and why
This notice applies to personal information collected through delphic.services. Delphic is a trading name used by Tarak Batra (“Delphic”, “we” or “us”). Tarak Batra determines why and how the personal information described here is processed and is the controller or Data Fiduciary responsible for it. Questions and rights requests can be sent to tarak@delphic.services.
Information you give us
When you submit a consultation or Knowledge Index form, we collect your work email and, depending on the form, your company website and optional phone number. We also record the page the request came from, the type of request and the time it was submitted. If you contact us another way, we use the information contained in that correspondence.
Information collected when you use the site
Our hosting and security systems may process your IP address, request time, requested page, browser or user-agent information and related diagnostic data.
Our site analytics can also record the page URL and title, referrer, campaign parameters, a randomly generated visitor and session identifier, language, timezone, device and viewport dimensions, page navigation, time on page, scroll milestones, selected link and button interactions, outbound links and source-reference clicks. The tracker does not record the text you enter into Delphic’s forms or Knowledge Index search.
When Ask Delphic is available, the optional analytics tracker remains unloaded for that page session so chat activity and the analytics identifier are not processed together. If the site cannot determine whether chat is available, the tracker also remains unloaded.
The persistent visitor identifier is stored in your browser until you clear the site’s stored data or we replace the identifier. Session information is held in session storage and a new session is created after 30 minutes of inactivity.
Questions entered into Knowledge Index search
When you use Search our knowledge, the question is sent to Delphic’s search function hosted by Vercel and compared deterministically with the public Knowledge Index. The search does not send the question to an external AI provider and does not generate a new answer from private Delphic knowledge.
Delphic does not currently save the exact question in its application database or website analytics. Operational logs record limited facts such as question length, search mode, number of results and processing time, without the question text. Do not enter confidential, commercially sensitive or unnecessary personal information into the search box.
Questions entered into Ask Delphic
When the public chatbot is enabled and you ask it a question, Delphic’s Vercel-hosted chat service validates the request and selects up to two relevant excerpts from the published Knowledge Index. It then sends your question, those bounded public excerpts and up to the three most recent successfully grounded turns in the current browser session through Cloudflare AI Gateway to Cloudflare Workers AI to produce a sourced answer. Questions rejected by the safety boundary and deterministic contact, legal or out-of-scope responses are not added to model history. The chatbot has no access to Delphic’s GitHub repository, private client knowledge, unpublished material, credentials or deployment configuration.
Delphic does not keep a server-side transcript or create a chatbot account. An encrypted and authenticated session value containing the recent turns and question count is returned to your browser and held only in page memory; refreshing or closing the page clears it. Application logs omit the question, answer and encrypted state. They may record lengths, result mode, public source identifiers, corpus version, token counts, response time and a sanitised upstream status. Cloudflare payload logging is disabled so its AI Gateway retains metadata rather than prompt or answer content. Cloudflare states that Workers AI customer content is not used to train models without consent.
Cloudflare Turnstile and rate limits are used as essential security measures to prevent automated abuse of the chatbot. Turnstile runs a browser challenge and validates a short-lived response token before an AI request is made; Cloudflare may process request, browser and network signals for that purpose. Do not enter personal, confidential, commercially sensitive, client-specific or credential information into the chatbot.
Requests to the public Knowledge MCP
The MCP serves the same published Knowledge Index material available on this website. Its application logs may record question length, the tool used, public record identifiers returned or fetched, result count, status, response time, corpus version, country or region, origin or referrer class, primary language and the lengths or types of supplied client and JSON-RPC identifiers.
Delphic does not intentionally retain raw MCP search text, raw JSON-RPC request bodies, JSON-RPC identifiers, client-supplied names or versions, fetched article content, authorisation or cookie headers, raw IP addresses, full user-agent strings, precise location, or full origin and referrer URLs in MCP application logs. The service creates no conversation history. Do not submit confidential, commercially sensitive or unnecessary personal information in an MCP search question.
Automated and AI crawler requests
When a request appears to come from an automated AI or search crawler, we may send the requested Delphic URL, request time, request method, IP address and user agent to our referral-measurement provider. This helps us understand whether Delphic’s published Answer Objects are being discovered by automated systems. Delphic uses the resulting reporting at an aggregated page level.
Our purposes and legal bases
- Responding to you. We use enquiry information to review and answer your request and, where relevant, to take steps towards a possible engagement at your request.
- Operating and protecting the website. We process technical information to deliver pages, maintain security, prevent misuse and diagnose faults. We do this for our legitimate interests in running a reliable and secure service.
- Finding and explaining relevant public knowledge. We process search and chatbot questions to route you to, or produce a sourced explanation from, published pages. We do this at your request and for our legitimate interest in making the Knowledge Index useful and navigable.
- Understanding discovery and interest. We use analytics and referral information to understand how people and automated systems reach and use our work, and to improve the site. Where applicable law requires consent for browser storage or analytics, consent is the basis for that activity. Otherwise, we rely on our legitimate interest in understanding and improving the effectiveness of our website.
- Meeting legal obligations. We may preserve or disclose information where the law requires it, or where necessary to establish, exercise or defend legal claims.
Where we rely on legitimate interests, we consider the effect on the people concerned and limit the information to what is proportionate to the purpose.
Who receives the information
Personal information is available only to people and providers who need it for the purposes above. These may include:
- Delphic team members responsible for enquiries, website operations and analysis;
- Vercel, which hosts the website and chat service and stores consultation records in private Blob storage;
- Cloudflare, which verifies chatbot requests through Turnstile and processes bounded chatbot questions and published excerpts through AI Gateway and Cloudflare Workers AI when the chatbot is enabled;
- our configured notification service, which receives only an opaque private-record identifier, request type and submission time, without your email, website, phone number or form origin;
- our website and AI-referral measurement provider, currently reached through events.peach.study;
- professional advisers, regulators, courts or public authorities where reasonably necessary or legally required.
We do not sell personal information or provide it to third parties for their own direct marketing.
International processing
Some of our service providers may process information in countries other than the one from which you visit. Where data-protection law requires a transfer safeguard, we use an applicable mechanism such as an adequacy decision or approved contractual protections provided by the relevant service provider.
How long we keep it
- Consultation and Knowledge Index enquiries: no more than 180 days in the private enquiry store, unless an engagement begins or we need a limited record for a legal or administrative reason. The cleanup process uses a 179-day deletion threshold as an operating margin. The notification workflow receives no contact details. Information required for an engagement is then handled under the engagement terms and the applicable retention requirements.
- Browser identifiers: the persistent identifier remains until cleared or replaced; session information is renewed after 30 minutes of inactivity and is removed when the browser session storage is cleared.
- Knowledge Index search: Delphic does not currently retain the exact question in its analytics or application database. Limited operational logs omit the text, and the search does not send the question to an external AI provider.
- Ask Delphic: the current conversation and six-question count remain in encrypted and authenticated browser memory for the page session and are not stored in a Delphic transcript database. Chat-service application logs and Cloudflare AI Gateway metadata follow the retention settings of the relevant hosting accounts, but are configured to omit prompt and answer payloads.
- Knowledge MCP: application logs omit raw search text, raw IP addresses, full user agents, precise location, protocol bodies, authorisation and cookie headers, and returned article content. Limited operational events are retained only as long as needed to operate, secure and assess the public service.
- Analytics, crawler and security information: only for as long as needed to produce aggregate reporting, operate and secure the service, investigate a specific issue or meet a legal requirement. Aggregated information that no longer identifies a person may be retained for longer.
We may delete information earlier when it is no longer useful for the purpose for which it was collected.
Security
We use access controls, private storage, restricted administration and transport security intended to protect personal information. No internet service can guarantee absolute security, but we limit access and review the safeguards appropriate to the information we hold.
Your choices and rights
Depending on where you live, you may have the right to ask for a copy of your personal information, have inaccurate information corrected, have information deleted, restrict or object to processing, or receive certain information in a portable form. Where processing depends on consent, you may withdraw that consent at any time without affecting earlier lawful processing.
Send a request to tarak@delphic.services. We may need enough information to verify your identity and locate the relevant record. You may also complain to the data-protection authority where you live or work if you believe your information has been handled unlawfully.
Client knowledge and engagement data
This website notice does not replace the confidentiality, data-processing, security and retention terms agreed for client work. A client’s research, data, methods, deliverables and private knowledge are handled under the relevant engagement agreement.
Changes to this notice
We may update this notice when the website, our providers or legal requirements change. The date above identifies the current version. Material changes will be made clear on this page.
PART 02 / WEBSITE TERMS
The terms for using this website
These terms are between you and Tarak Batra, trading as Delphic, and apply when you use delphic.services. By continuing to use the website, you agree to use it lawfully and in accordance with these terms. A paid or prospective client engagement is governed by its own written agreement, not by these website terms.
Information, not professional advice
The website explains Delphic’s point of view, methods and services. Its content is general information, not legal, financial, investment or other regulated professional advice. It is not a proposal or a substitute for advice based on your particular circumstances.
No promised AI outcome
AI systems, search products and source environments change and can produce variable results. Unless Delphic expressly agrees otherwise in writing, nothing on this website guarantees that a person, firm or page will be retrieved, cited, credited, ranked or recommended by any particular system.
Our content and intellectual property
Unless stated otherwise, Tarak Batra owns or is licensed to use the website, its design and its original content. You may view it, link to it and quote reasonable portions with clear attribution. Ordinary indexing, retrieval and citation by search and AI systems are permitted where consistent with our published access controls.
You may not remove attribution, misrepresent Delphic’s work, reproduce a substantial part of the site as your own, use automated access in a way that disrupts the service, or systematically extract the content to create a competing publication, dataset or service without written permission.
Sources and third-party websites
Answer Objects may cite external research and link to third-party websites. Those sources remain responsible for their own content and availability. A link or citation does not mean that Delphic controls or endorses everything published by that third party.
Availability and changes
We may correct, update, withdraw or reorganise website content and may suspend access for maintenance, security or operational reasons. We aim to keep the website useful and accurate but do not promise that every page will always be available, complete or current.
Knowledge Index search
Search is a navigation aid over Delphic’s public material. Results may be incomplete or imperfect and do not disclose protected knowledge merely because a related capability is listed. You must not submit confidential, commercially sensitive or unlawful material, probe the service to recover protected information, or use automated requests in a way that disrupts or burdens it.
Ask Delphic chatbot
The chatbot is an automated guide to Delphic’s published Knowledge Index, not professional advice or a statement tailored to your circumstances. It can be incomplete, unavailable or wrong despite its source controls. Check the linked published pages before relying on an answer. You must not use the chatbot to seek private or unpublished information, extract protected methods, bypass safeguards, submit personal or confidential material, or automate requests. The six-question session limit may be changed or access may be suspended to protect the service. Chatbot output does not promise an AI, commercial or engagement outcome.
Responsibility for loss
To the fullest extent permitted by applicable law, Delphic is not responsible for indirect or consequential loss arising from reliance on this website or inability to use it. Nothing in these terms excludes responsibility that cannot lawfully be excluded, including responsibility for fraud or fraudulent misrepresentation.
Privacy and changes to these terms
The Privacy Notice above explains how information connected with website use is handled. We may update these terms as the website changes. The version in force is the one published here with the current update date.
Questions and permissions
For privacy questions or permission to reuse Delphic content beyond the allowance above, contact tarak@delphic.services.
Separate engagement terms apply. Nothing on this page changes a signed agreement with Delphic or makes private client information part of this public website.